Cache existing chunks in frame 0.

[joaander]

Description

Cache the results of chunk_exists calls from frame 0.

Motivation and Context

This increases performance. In cases where users write many values that match defaults, the previous code would check (for every single chunk) whether that chunk is present in frame 0:

DEBUG:gsd.hoomd:skipping data chunk, matches frame 0: particles/orientation
DEBUG:gsd.fl:chunk exists: test.gsd - particles/velocity
DEBUG:gsd.hoomd:writing data chunk: particles/velocity
DEBUG:gsd.fl:write chunk: test.gsd - particles/velocity

The chunk_exists call triggers an implicit flush which decreases write performance significantly.

After this change, the chunk_exists call is made once for frame 0 and the results are reused, resulting in no extra chunk_exists calls:

DEBUG:gsd.hoomd:skipping data chunk, matches frame 0: particles/orientation
DEBUG:gsd.hoomd:writing data chunk: particles/velocity
DEBUG:gsd.fl:write chunk: test.gsd - particles/velocity

How Has This Been Tested?

Existing tests pass.

hoomd-benchmarks.py does not trigger the extra chunk_exists calls, so it does not show any performance improvement. Use-cases that trigger it excessively should see improvement.

Checklist:

• I have reviewed the Contributor Guidelines.
• I agree with the terms of the GSD Contributor Agreement.
• My name is on the list of contributors (doc/credits.rst) in the pull request source branch.
• I have added a change log entry to CHANGELOG.rst.