[python] Remove double open For SOMAArray reads

[nguyenv]

Issue and/or context:

This finishes up #3053 [sc-54744]

Changes:

• Remove clib.SOMAArray.open in all SOMAArray derived class read calls
• Pass query arguments coords, column names, result order, value filter, and platform config that were formerly passed to clib.SOMAArray.open or clib.SOMAArray.reset instead to TableReadIter and SparseNDArrayRead
• _set_coords on the ManagedQuery rather than through SOMAArray
• _arrow_table_reader and SparseTensorReadIterBase initialize a ManagedQuery object and set the query arguments listed above; set up the read; and submit reads and return results until the query is complete

[codecov]

Codecov Report

Attention: Patch coverage is 94.04762% with 5 lines in your changes missing coverage. Please review.

Project coverage is 85.32%. Comparing base (d0cbd2d) to head (9124aff).
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3293      +/-   ##
==========================================
+ Coverage   85.09%   85.32%   +0.22%     
==========================================
  Files          53       53              
  Lines        5577     5567      -10     
==========================================
+ Hits         4746     4750       +4     
+ Misses        831      817      -14     

Flag	Coverage Δ
python	85.32% <94.04%> (+0.22%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
python_api	85.32% <94.04%> (+0.22%)	⬆️
libtiledbsoma	∅ <ø> (∅)

[johnkerl]

This is really really nice --a well-thought-out refactor, well executed. Just some minor mods requested. :)

[johnkerl]

🚢

[bkmartinjr]

This looks great. Two small items:

1. platform_config handling - noted inline
2. ManagedQuery GIL - the read path moves some logic up from C++ to Python -- we need to make sure the pybind entry points are all releasing GIL where appropriate. My quick skim is that it looks OK, but I suggest we benchmark after landing this to confirm.