Post-workshop improvements

[MBartkowiakSTFC]

Description of work The recent live demonstration of MDANSE has uncovered several unrelated minor problems. A few of them are addressed here.

Fixes

1. ASE converter will now reuse the initial unit cell definition if the input file (tried on a PDB file) only contains the definition for the first frame,
2. TrajectoryWriter is now able to resize the datasets to the number of frames that were successfully read. This is at the moment only used by the ASE converter (which is most likely to predict the number of frames wrong).
3. Some error messages have been relabelled as warnings, most notably the missing configuration file messages,
4. StructuredSession has been modified to reduce the likelihood of it keeping many files opened at the same time.

To test All tests must pass. Please try to convert an openmm PDB trajectory using ASE converter. The resulting trajectory should have the correct number of frames and a unit cell defined at all steps.

[MBartkowiakSTFC]

The file fails because it has more molecules than ASE can handle.

I created an issue here: https://gitlab.com/ase/ase/-/issues/1539 to see if anything can be done about it on the ASE side.