feat(firestore): add retry support for WriteBatch commit

[developerjamiu]

Related Issues

Fixes #63

Problem

Standalone write operations (set, update, create, delete) use WriteBatch._commit() to send a single HTTP request to Firestore. If the server returns a transient error such as UNAVAILABLE, RESOURCE_EXHAUSTED, or ABORTED, the operation fails immediately with no retry — even though the same request sent moments later would succeed.

The retry infrastructure (ExponentialBackoff, commitRetryCodes) was already ported from the Node.js SDK but never wired into the commit path.

Solution

Added a retry loop to WriteBatch._commit() using the existing ExponentialBackoff and _maybeBackoff() utilities, matching the pattern already used in _runTransaction().

Retry codes follow the Node.js SDK (write-batch.ts line 443): ABORTED + commitRetryCodes (RESOURCE_EXHAUSTED, UNAVAILABLE).

Transaction commits are excluded from this retry loop — they pass a transactionId and are already retried at the transaction layer.

Changes

• write_batch.dart: retry loop in _commit() for standalone commits
• write_batch_test.dart: 5 tests covering retry and non-retry paths

Checklist

Before you create this PR confirm that it meets all requirements listed below by checking the relevant checkboxes ([x]).

• I have updated the CHANGELOG.md of the relevant packages.
  Changelog files must be edited under the form:

## Unreleased minor

- Added retry support for `WriteBatch.commit()` on transient errors (`ABORTED`, `UNAVAILABLE`, `RESOURCE_EXHAUSTED`). (thanks to @developerjamiu)

• If this contains new features or behavior changes,
  I have updated the documentation to match those changes.