fix(file_store): don't overwrite entries appended via another handle

[Brijesh-Thakkar]

Description

Fixes #2308.

Store::append writes at the file handle's own cursor, and the file is not opened in append mode. That cursor is only synced on load, dump and the handle's own appends. If another handle to the same file appends in the meantime, an append through the stale handle overwrites that entry. When the entries are the same size the file stays decodable, so the earlier changeset is lost silently.

Before writing, append now scans forward from the handle's cursor using the existing EntryIter, skipping valid entries that other handles added. The cursor ends up right after the last valid entry, and the new changeset is written there.

Notes to the reviewers

• Replace bincode by postcard #2258 addresses the same issue by seeking to the end of the file before each write. As raised in review there, that makes a corrupted tail unrecoverable, because later entries land behind undecodable data. This PR keeps the current "append only after valid data" behavior: it reuses EntryIter and writes where valid data ends. A test covers appending after a failed dump and after a partial trailing entry.
• Errors during the scan are handled as follows:
  • Clean EOF ends the scan and the write happens at the cursor.
  • Garbage or a partial entry (UnexpectedEof) ends the scan and the write starts there, as on master.
  • Real I/O errors are returned instead of ignored.
• Single-handle behavior is unchanged. The cursor is already at EOF, so the scan costs one extra read and the write lands where it always did.
• No new dependencies, fields or public API changes. Length framing and the set_len rollback from Replace bincode by postcard #2258 are intentionally out of scope.
• Limitations:
  • Concurrent writers still race. Two handles writing at the same moment can interleave or clobber each other. This change only fixes appends made after another handle has finished writing.
  • A lockfile or a one-handle-per-file design remains the real long-term answer.
  • EntryIter::drop ignores a failed cursor-sync seek, so that one seek failure can go unnoticed.

Changelog notice

Fixed

• bdk_file_store: Store::append no longer overwrites changesets that were appended through another handle to the same file.

Checklists

All Submissions:

• I followed the contribution guidelines

Bugfixes:

• This pull request breaks the existing API
• I've added tests to reproduce the issue which are now passing
• I'm linking the issue being fixed by this PR

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[nymius]

NACK 7b6f7f8

bincode has been deprecated. Any solution for the file store issues should also carry the migration from bincode to postcard.