ASV benchmarks for bool column truncation

[IvoDD]

Reference Issues/PRs

New benchmarks for memcpy bool truncation.
Useful for benchmarking follow up PR #3052

What does this implement or fix?

Any other comments?

Checklist

Checklist for code changes...

• Have you updated the relevant docstrings, documentation and copyright notice?
• Is this contribution tested against all ArcticDB's features?
• Do all exceptions introduced raise appropriate error messages?
• Are API changes highlighted in the PR description?
• Is the PR labelled as enhancement or bug so it appears in autogenerated release notes?

[claude]

Consider adding peakmem_read_row_range_byte_aligned and peakmem_read_row_range_byte_unaligned counterparts to match the existing time_*/peakmem_* pairing pattern used for time_read/peakmem_read and time_write/peakmem_write. Memory usage during truncation may differ meaningfully between aligned and unaligned cases.

[claude]

ArcticDB Code Review Summary

API & Compatibility

• No breaking changes to public Python API
• On-disk format unchanged
• No protobuf or key type changes

Memory & Safety

• No C++ changes; not applicable

Correctness

• Row range values are correct: (8, 100_008) produces byte-aligned offsets (8 is divisible by 8) within each 100K-row block; (5, 100_005) produces byte-unaligned offsets (5 is not divisible by 8)
• Comments accurately explain the intent of the specific values

Code Quality

• Missing peakmem_read_row_range_byte_aligned / peakmem_read_row_range_byte_unaligned counterparts -- the class uses time_*/peakmem_* pairs throughout; memory profiling during truncation may differ between aligned and unaligned cases (inline comment posted)

Testing

• This PR is the test/benchmark addition; no behavioural code changed

Build & Dependencies

• No CMake or dependency changes

Security

• No credentials or boundary violations

PR Title & Description

• Title is clear and concise
• Description wording is confusing: Useful for benchmarking follow up PR 3052 -- since PR 3052 has a lower number it presumably predates this PR; clarify whether these benchmarks measure the already-merged 3052 work or whether 3052 is forthcoming and depends on these benchmarks landing first
• PR checklist items left unchecked (likely fine for a benchmarks-only PR, but worth acknowledging)

Documentation

• No public API changes; no documentation updates needed

---

Side-effect note (informational): The setup_cache_key values for ArrowStrings benchmarks changed from arrow:263 to arrow:271 in benchmarks.json because inserting 8 new lines into ArrowBools shifted those line numbers. This is correct ASV behaviour and will simply invalidate the cached setup for those benchmarks on the next run -- no action needed.