Nestcheck

Changed

* Now compatable with latest version of fgivenx (v2.1.11).
* Minor refactoring of efficiency gain calculations.
* Adding warnings for a bug which occurs in output files from PolyChord v1.13.
* Minor corrections of typos and references in paper.md.

Added

* When analytic results are available, implementation error calculations can now use RMSEs as well as the standard deviation of results. This takes account of any systematic bias in results.
* Option to turn off duplicate logl warnings.
* "thin" and "point size" arguments in logx diagram plotting.
* Random seeding is now used when dealing with duplicate logl values, so results are always reproducible.

Changed

* Minor improvements to documentation.

Changed
* duplicate logl values in data files are now allowed (a unique order is assigned in get_birth_inds)
* several minor bug fixes

Removed:
* `check_logls_unique` function in data processing (no longer needed)

Added:
* `check_logls_unique` function in data processing which checks if any point loglikelihoods are duplicates and, if so, raises a UserWarning and then adds a tiny random delta to each duplicate. This allows a unique point likelihood order for calculating threads - previously duplicates would have resulted in an error.

Removed:
* symlink from examples to demos (caused problems with codeclimate).

Added:
* `get_w_rel` helper function in `ns_run_utils`.
Changed:
* Improved `summary_df_from_multi` to fix multi-indexing edge case which was causing problems with the new version of `pandas` (v0.23.0).

Added:
* `process_dynesty_run` function in `nestcheck.data_processing` for loading `dynesty` results into nestcheck format. The tests were also updated according.