Mrc

--------------------------

* Add helper functions for working with header labels.
* Update header statistics calculation to use float32 instead of float64. This
is about twice as fast and uses less memory, but does make the calculations
slightly less accurate and will overflow (to 'inf') if the data array
contains very large values (larger than approximately 1e19).

--------------------------

* Add support for passing file names as pathlib Path objects.

--------------------------

* Add file names to mrcfile-header output
* Improvements to validation functions (mrcfile-validate command and the
mrcfile.validate() function):
* Add a message with the file name of each file processed, to enable valid
and invalid files to be identified.
* Add a message confirming that valid files are valid.
* Ensure that all files are processed when validating multiple files, even
if some are invalid or raise exceptions on opening.
* Add support for numpy 1.23 and Python 3.10 and 3.11

--------------------------

* Change default MRC format version from 20140 to 20141 and update validation
to allow files with both format numbers as valid.
* Add mrcfile.read() and mrcfile.write() functions for quick one-off accesses
to the MRC data array without needing to keep an MrcFile object open (thanks
to Alister Burt).
* Allow the extended header to be set when creating a file with
mrcfile.new_mmap() (thanks to James Parkhurst).
* Add file size checks to avoid allocating huge amounts of memory when opening
corrupt or non-MRC files in permissive mode.
* Update numpy and Python version compatibility to drop support for Python
3.4 and 3.5 and numpy < 1.16, and add support for numpy 1.21 and 1.22.
* Fix bug intepreting FEI extended headers on big-endian hardware.
* Documentation updates.

--------------------------

* Add support for the new proposed data mode 12 (numpy float16). Note that
this causes a change in behaviour: previously float16 arrays would be
widened to float32 and saved in mode 2; now they will be kept in float16 and
saved in mode 12, which could be incompatible with other software.
* Issue Python warnings if NaN or infinite values are found when a new data
array is passed to set_data() (or if update_header_stats() is called).
* Update numpy version compatibility to include numpy 1.20
* Minor updates to documentation