chore: improve fastwebsockets maintenance path

[oab24413gmai]

Summary:

• Add edge-case unit tests in src/mask.rs for the unmask routine covering zero-length payloads, length-1, length-3 (sub-word), and length spanning a 4-byte boundary plus tail bytes, asserting round-trip mask/unmask equivalence and that the mask key indexing matches RFC 6455 §5.3. Optionally add a frame-roundtrip test in src/frame.rs for a payload at the 125/126 length boundary to exercise the 16-bit length encoding path.
• Keep the change narrow so it is straightforward to review.

Notes:

• I kept this scoped to the relevant implementation and tests.

[divybot]

Maintenance scan note: CI is failing only at the formatting step after build and tests passed. The workflow runs cargo fmt -- --check --verbose on all platforms, so this should just need cargo fmt on the branch and a push.

Once formatting is fixed, this looks like a narrow test-only PR; no need to rerun anything broader than the existing CI unless the patch changes beyond formatting.

[divybot]

Maintenance scan note: CI failed on the Check formatting step. cargo fmt --check reports three diffs in the new tests in src/frame.rs:

• around line 385: the 127 => arm's let len = u64::from_be_bytes(buf[2..10].try_into().unwrap()) as usize; should fit on one line.
• around line 393: the mask-key array should be vertical-formatted —

  let m = [
    buf[offset],
    buf[offset + 1],
    buf[offset + 2],
    buf[offset + 3],
  ];

• around line 414: let (fin, opcode, mask, payload_len, offset) = parse_frame_header(&buf); should fit on one line.

Running cargo fmt locally and pushing should clear all three jobs. Ubuntu and Windows were cancelled by the macOS failure rather than failing independently, so a clean cargo fmt is likely the entire fix.

The test shapes themselves — 125/126 length-encoding boundary and the length-0/1/3/7 mask cases — look like reasonable gap-fillers.

[littledivy]

Thank you, we have plenty tests and this PR fixes no bug