Updated pargen.utils.load_data to use LH5Iterator and field_mask to be more memory efficient

[iguinn]

load_data was using too much memory (at least in P08 which is a larger dataset), and crashing my attempts to process it on NERSC. Switch it to use the fieldmask when reading from the input files and the LH5Iterator to limit the number of entries in memory at once. Also improved commenting/docstring.

Note this should not be merged until this is also merged: legend-exp/legend-pydataobj#100

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 29 lines in your changes missing coverage. Please review.

Project coverage is 48.91%. Comparing base (981877e) to head (ee73d2a).

Files	Patch %	Lines
src/pygama/pargen/utils.py	0.00%	29 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #589      +/-   ##
==========================================
+ Coverage   48.80%   48.91%   +0.11%     
==========================================
  Files          59       59              
  Lines        7846     7821      -25     
==========================================
- Hits         3829     3826       -3     
+ Misses       4017     3995      -22     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[gipert]

Is this tested @ggmarshall? @iguinn can you bump the pydataobj version in pyproject.toml, if this is not backward compatible?

[ggmarshall]

Not on my end but I can have a look later this week