docs(batch): drop restatement comments in tests and sources

[oferchen]

Summary

Per-crate comment cleanup pass on the batch crate, following the precedent set by PRs #4587 (matching), #4591 (protocol), #4590 (transfer), #4589 (fast_io), #4584 (filters), #4582 (logging), #4583 (platform), #4575 (compress).

The crate already enforces #![deny(missing_docs)] at the root, so every public item already carries ///. This follow-up strips pure restatement and decorative section labels from the test bodies and a handful of source files, while preserving every upstream rsync citation, batch-shell-script generation WHY note, and load-bearing protocol invariant.

Scope

Walked crates/batch/src/ end to end. Files touched: 7.

• src/tests.rs - integration tests: drop // Write a batch file, // Read the batch file back, // Verify file entries round-trip correctly, // EOF, // Should fail on corrupt data, // File 1 entry/// File 2 entry, per-section // Directory entry/// File entry/// End of flist labels, // File checksum (16 zero bytes) decorators, and // Replay the ... batch markers above the obvious replay call.
• src/reader/tests.rs - drop // Try to read more data than available, // Write phase/// Read phase decorative dividers, // Don't read header, and a cluster of // Initial segment, // Read the sub-list segment, // Now we have all 3 entries, // NDX codec is available for ..., // Sub-list entries have parent directory prepended labels in the INC_RECURSE test builder.
• src/format/tests.rs - drop value-restating comments above assert_eq!(&buf[N..M], ...) byte-layout checks; rewrite the multi-byte varint length breakdown as a single concise line.
• src/replay/tests.rs - drop // Basis has "ABCDE" and // sqrt(1_000_000) = 1000 type restatements; convert the two CPRES_ZLIB / zstd cpres_zlib_* tests' explanatory bodies into /// rustdoc on the test functions themselves (where they actually belong).
• src/reader/delta.rs - tighten the simple_recv_token match-arm comments (block_index formula stays, redundant "Literal data: n bytes follow" trailing the same-line literal_len reads).
• src/script.rs - drop // Original filter args should be stripped, // The batch name should not appear as a passthrough argument, and the redundant // Read and verify script content -- upstream has no shebang (kept as the WHY note upstream batch shell scripts have no shebang).
• src/writer.rs - drop // Verify file exists and has content and // write_data before write_header must fail decorators above the corresponding assert!.

What was kept

• All // upstream: batch.c:NNN, io.c:NNN, compat.c:NNN, flist.c:NNN, token.c:NNN, uidlist.c:NNN, main.c:NNN, receiver.c:NNN, rsync.c:NNN citations.
• All WHY notes: do_compression flag semantics, batch body being a raw protocol-stream tee, NDX codec state preservation across flist sub-lists, INC_RECURSE ndx_start segment layout, CPRES_ZLIB dictionary sync via see_token() after block matches, zstd codec auto-detection, MD4/MD5/XXH3 default 16-byte digest length rationale, iflags: ITEM_TRANSFER (0x8000) annotation above the raw u16 byte writes (the literal does not document itself).
• All non-trivial test docstrings (/// Verifies that ...) including the upstream-bug reproductions for token.c:608 inflate -3.

Stats

• 7 files changed, +47 / -234.
• 175 pure-comment lines removed; the remaining diff covers reflowing a few multi-line comment paragraphs into concise one-liners and moving two test-body explanations into /// rustdoc on the test fn.
• No semantic changes. No new #[allow(...)] attributes. No public-API rustdoc additions were needed (the crate enforces missing_docs).

Test plan

• cargo fmt --all -- --check
• cargo clippy --workspace --all-targets --all-features --no-deps -- -D warnings
• CI matrix (fmt+clippy, nextest stable, Windows, macOS, Linux musl)