[c++] Provide different modes of handling memory when converting to Arrow objects

[XanthosXanthopoulos]

Issue and/or context:
#4299 #4311 provide memory optimization to reduce overall memory copies and overallocation of memory but introduced a necessary memory copy in the read path.

Changes:
This PR introduces the option to optimize memory management for read operations based on two modes PERFORMANCE and EFFICIENCY

• PERFORMANCE: The allocated buffers used for reading from TileDB are passed to the Arrow arrays and a new set of buffers is allocated for subsequent reads. This mode preserves the current behavior but can be wasteful on memory
• EFFICIENCY: A new set of buffers is allocated to be used by the Arrow array. These buffers are appropriately sized to exactly match the data read from TileDB. A memory copy operation is then performed for each buffer. If the read buffers are completely filled then the memory copy is avoided and the buffers are used directly by Arrow and a new buffers is allocated for TileDB reads.

Notes for Reviewer:
There are currently two almost equivalent implementations for ColumnBuffer to be used during reads. The std::vector one uses a custom allocator to avoid initializing the memory during resize. Under Debug the custom allocator is not being optimized by the compiler and the resulting code is orders of magnitude slower (this can be somewhat mitigated by using -Og instead of -O0 optimization). In Release the compiler generated almost identical code with the array based implementation.

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.83%. Comparing base (1a00005) to head (5a81aba).
⚠️ Report is 1 commits behind head on xan/SOMA-688.

Additional details and impacted files

@@               Coverage Diff                @@
##           xan/SOMA-688    #4334      +/-   ##
================================================
- Coverage         86.83%   86.83%   -0.01%     
================================================
  Files               138      138              
  Lines             20743    20751       +8     
  Branches             16       16              
================================================
+ Hits              18013    18019       +6     
- Misses             2730     2732       +2     

Flag	Coverage Δ
python	89.15% <80.00%> (-0.02%)	⬇️
r	85.61% <100.00%> (-0.01%)	⬇️

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

Components	Coverage Δ
python_api	89.15% <80.00%> (-0.02%)	⬇️
libtiledbsoma	76.77% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.