Viperleed

Only patches tleedm. See viperleed-betatest repository for executable. Skipping number 0.5.5, which was a temporary local patch but not uploaded to Github.

- Fixed several major bugs in domain calculations. Most of these bugs were program-stopping, so any domain calculations that were produced with previous versions should be fine if the execution finished without errors.
- Fixed a bug with symmetrical displacements in four-fold symmetric systems, caused by a sign error in rotation matrices. This only affected the displacement vectors, not the symmetry detection itself. Refactored code to avoid similar errors in the future, all rotation matrices now access the same definition.
- Fixed a minor issue with search behaviour when the GAUSSIAN_WIDTH parameter is already at the minimum (due to fortran input precision). This is mostly cosmetic, as the search will still stop and restart to modify the population according to SEARCH_CULL.

Only patches tleedm. See viperleed-betatest repository for executable.

- Reversed the order of energy steps for parallelized reference calculations. This makes more efficient use of the available CPUs, because the faster low-energy calculation now come last, which makes having to wait for one core that is slow to finish less likely.
- Added a warning if symmetry-equivalent atoms are assigned different site types.
- Fixed a bug concerning ASBULK/ASA bulk repeat vectors in reference calculation input. This concerned systems with two inequivalent bulk layers and bulk repeat units that are *not* perpendicular to the surface.
- Fixed a bug in writing rf.info: fortran *does not* require a line break after the 25I3 fields.
- Fixed a bug in creation of POSCAR_bulk_appended that caused duplicates of bulk atoms to be created if multiple bulk layers were added.

Only patches tleedm. See viperleed-betatest repository for executable.

- added a job.py script that replaces job.sh when running directly from python
- set LOG_SEARCH default to True (note: this may create large files for TensErLEED versions below 1.61)
- fixed a bug in reading of Woods notation, relevant for e.g. SUPERLATTICE
- fixed several minor bugs concerning "domain" calculations
- fixed bug that would crash the reference calculation if TENSOR_OUTPUT was set to False
- fixed a bug in bulk symmetry detection (always needs to search for unit cell origin)
- fixed a formatting issue that would make some refcalc-fd.out files generated from parallel reference calculations unusuable for R-factor calculations

Also some major refactoring of tleedm, and new functionality for interfacing with ase.atoms

Only patches tleedm. See viperleed-betatest repository for executable.

- added plotting of phaseshifts as Phaseshifts_plots.pdf at initialization
- fixed a bug that caused tleedm to crash after the search if no R-factor data had been read yet
- introduced more robust handling of mpirun subprocess children in the search to avoid zombie processes when the search is stopped early
- refactored detection of "surface atoms", slightly increased atom radius that is considered to "cover" underlying atoms from 1.1 times covalent radius to 1.2 times c.r.
- fixed a minor bug that would initialize search parameters concerning vacancies at the same value as the main site element, even though they actually only have one "step". TensErLEED was apparently robust to this, did not seem to cause any issues.
- added edge texts for occupation parameters in Search-progress.pdf and Search-report.pdf

Only patches tleedm. See viperleed-betatest repository for executable.

- Added option for minus signs / overbars in PLOT_RFACTOR parameter
- Fixed some minor bugs in reading EXPBEAMS.csv: Accept files without .csv extension, accept float values, check for data after reading
- Made fortran compile statements more error-tolerant: A typical problem in compilation is that the '-mcmodel=medium' flag is missing. If fortran compilation fails and the fortran compile log indicates that this is the reason, tleedm will now warn, but try to recompile, adding the '-mcmodel=medium' flag. If this still fails, execution stops as usual.
- Fixed a bug that prevented THEOBEAMS from being written after the reference calculation
- Made VIBROCC_OUT always write the elements explicitly in mixed-occupation sites, even if only one element remains after the search
- Fixed a bug in error calculations that led to output not being written if the number of displacements steps was not the same for all atoms
- Fixed the definition of var(R) in error calculations, which was off by a factor of min(R)
- Improved plotting after error calculation: cleaned up spline fit, better positioning of the "R + var(R)" labels.

Please note: This release is backwards compatible with TensErLEED version 1.6, but if you are using the (partly modified) TensErLEED version 1.61, you *must* update the lib/lib.delta and src/delta files in the tensorleed/TensErLEED-v1.61 directory, otherwise tleedm will crash.

**Feature changes:**
- *tleedm*: Parallelized reference calculations now use variable LMAX (depending on the energy), based on the PHASESHIFT_EPS parameter (see documentation). Requires TensErLEED version 1.61, constant LMAX will be used for older versions. This speeds up calculations especially at low energies. Note changed functionality of the LMAX parameter: Setting LMAX to a single value still fixes LMAX to this value for the entire energy range, disabling PHASESHIFT_EPS. However, passing *two* values to LMAX now defines upper and lower limits for LMAX, and PHASESHIFT_EPS is still used to determine appropriate LMAX values between these limits.
- *tleedm*: Added more formatting options for the Rfactor_plots.pdf output. The parameter PLOT_COLORS_RFACTOR does not exist any more, and the functionality is instead included with several other options (e.g. number of panels per page) in the PLOT_RFACTOR parameter. See documentation for details.
- *tleedm*: Improved memory handling of parabola fit: Reading in large data.chem files sometimes led to excessive memory usage, which could hamper performance. Now, the memory used for data in the parabola fit is limited to 1 GB (discarding data points with worst R-factors if they would exceed this limit), and is only stored temporarily.
- *tleedm*: Parabola fit now discards all data points with R-factors above min(R) + var(R). This, together with the point above, appears to somewhat improve the stability of the parabola fit.
- *tleedm*: Added option to disable the parabola fit entirely by setting 'PARABOLA_FIT = off' in the PARAMETERS file.
- *bookkeeper*: Added --discard (or -d) option. If the bookkeeper is called with ths option, it will try to delete all output from the last run (including a new Tensors.zip archive, if one was generated), not move anything to history, and not write to history.info.

**Bug fixes / minor changes:**
- *tleedm*: Fixed a bug that would reset SEARCH_START to 'crandom' after the first search, if multiple searches were queued.
- *tleedm*: Fixed a bug in the bulk-to-slab attachment interlayer vector for POSCAR files with off-normal c vectors (i.e. not perpendicular to the surface plane).
- *tleedm*: Improved handling of KeyboardInterrupt in parallel reference- and delta calculations. This should now always result in a clean exit, essentially the same as when setting STOP in the PARAMETERS file.
- *tleedm*: Changed y-axis range in Search-report.pdf