feat(zmodem): accelerate upload and download transfers

[lswzw]

直接粘贴以下内容：

## Summary

Implement ZMODEM upload and download acceleration for Netcatty by reducing per-byte array processing and payload allocations.

## Type of Change

- [ ] Bug fix
- [x] New feature
- [ ] Refactor / code cleanup
- [ ] Documentation update
- [ ] Build / CI change
- [ ] Other (please describe):

## Related Issue (optional)

N/A

## Changes Made

- Add Buffer/Uint8Array fast paths for ZMODEM receive and send data.
- Add typed-array CRC16/CRC32 processing and zero-copy download payload handling.
- Reduce upload allocation overhead with direct frame encoding and larger read chunks.
- Throttle upload progress events across transports.
- Improve ZMODEM send-session keepalive and handshake handling.
- Add `NETCATTY_ZMODEM_FAST_PATH=0` to disable both performance paths for troubleshooting.
- Add receive-path regression and protocol-equivalence tests.

## Screenshots / Demo

N/A - terminal transfer performance change.

## Testing

- [ ] I have tested these changes locally (`npm run dev`) (not applicable; no UI changes)
- [x] Linting passes (`npm run lint`)
- [ ] Tests pass (`npm test`) (targeted tests pass; full suite is blocked by the existing unmatched glob `electron/capabilities/*/*/*.test.cjs`)
- [x] Generated capability tool specs are updated when applicable (not applicable)
- [x] No new console errors or warnings, if this affects app behavior

Additional verification:

- Targeted ZMODEM and bridge tests: 4/4 passed.
- Receive parser benchmark: approximately 101 MB/s on local test data.
- `node --check` and staged diff checks passed.

## Checklist

- [x] My code follows the existing project style
- [x] I have added or updated relevant documentation (inline implementation comments)
- [x] I have not introduced any breaking changes (the fast receive payload is intentionally `Uint8Array` internally for zero-copy handling)