fix: use BitVec<u8, Lsb0> for platform-independent bit ordering

[eywalker]

Summary

• Validity bitmap: Changed from default BitVec (BitVec<usize, Lsb0>) to BitVec<u8, Lsb0> — fixes platform-dependent word size (usize is 4 bytes on 32-bit, 8 bytes on 64-bit)
• Boolean value packing: Changed from BitVec<u8, Msb0> to BitVec<u8, Lsb0> — consistent with Arrow's native bit ordering
• Bitmap length serialization: Cast len() to u64 before to_le_bytes() for stable 8-byte output regardless of platform

All bit vectors now use BitVec<u8, Lsb0>, matching Arrow's native validity bitmap layout.

Test plan

• All 32 lib tests pass (including updated byte-level boolean packing tests)
• All 19 integration tests pass (golden hash values updated)
• 13 ignored tests (pre-existing, for known unfixed bugs) unchanged

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR makes Arrow digest hashing deterministic across platforms by standardizing bitmap storage to BitVec<u8, Lsb0>, aligning boolean bit-packing with Arrow’s native LSB-first ordering, and ensuring bitmap length serialization is stable.

Changes:

• Switch validity bitmaps from default BitVec (usize-backed) to BitVec<u8, Lsb0> for platform-independent raw storage.
• Change boolean value packing from Msb0 to Lsb0 to match Arrow’s bit ordering.
• Serialize bitmap bit-length as u64 before hashing to ensure consistent 8-byte output.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/arrow_digester_core.rs	Standardizes bitmap representation and boolean packing; updates bitmap-length hashing for cross-platform stability.
tests/arrow_digester.rs	Updates golden digests to reflect the new stable bit ordering/serialization.

Comments suppressed due to low confidence (1)

src/arrow_digester_core.rs:338

• Boolean hashing for non-nullable arrays only feeds the packed bytes into the digest. Because the last byte is padded, different boolean lengths can produce identical byte streams (e.g., len=1 and len=8 with the same first bit), leading to guaranteed hash collisions. Prefix the boolean bit-length (or element count) into the digest before updating with bit_vec.as_raw_slice() so arrays with different lengths cannot collide.

                        for i in 0..bool_array.len() {
                            bit_vec.push(bool_array.value(i));
                        }

                        data_digest.update(bit_vec.as_raw_slice());

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

There are two extra blank lines before the use crate::arrow_digester_core::{...}; import; please remove them to keep the test module formatting consistent.

Suggested change

[Copilot]

Now that the validity bitmap uses BitVec<u8, Lsb0>, iterating each byte and calling to_be_bytes() is redundant and adds per-byte update overhead. You can update the digest in one call with the raw slice (and drop the clippy::big_endian_bytes expect) to keep this path simpler and faster.

[eywalker]

This is superseded by PR #9