This is an automated cherry-pick of #10607

What problem does this PR solve?

Issue Number: close #10606

Problem Summary:

1. During the BR restore process, TiDB first creates tables, performs region pre-splitting, and scatters the resulting regions across nodes. The regions created during this phase are empty.
2. TiFlash intentionally avoids creating IStorage objects when handling commands such as splits or snapshot applications for empty regions. This design choice helps reduce memory consumption and file fragmentation caused by empty tables in TiFlash.
3. However, in some cases, after a region is pre-split but before it finishes the PreHandleSnapshot stage, the region gets scheduled away (i.e., moved to another node). This triggers tiflash-proxy to call TiFlash’s releasePreHandledSnapshot.
4. Inside releasePreHandledSnapshot, the code retrieves an IStorage pointer from TMTStorages but fails to check whether it is a nullptr before directly invoking cleanPreIngestFiles. This missing null check leads to a null pointer dereference (NPE) and causes TiFlash to crash.

What is changed and how it works?

KVStore: ignore non-exist IStorage instance in releasePreHandledSnapshot
* `KVStore::releasePreHandledSnapshot` skip for external files is empty or IStorage instance is not exist

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Fix a issue that TiFlash may panic during br restore