Detect nested v1 filesystem data

[benthecarman]

FilesystemStoreV2 already rejected v1 data when a key file was found at the store root, but it did not inspect namespace directories. This missed v1 layouts such as primary/key, where v2 expects primary to contain secondary namespace directories.

For example, an ldk-node store can contain a BDK descriptor below a namespace directory. The previous check would accept that directory as v2 data because the root contained only directories, leaving the incompatible descriptor file undetected.

[ldk-reviews-bot]

👋 Thanks for assigning @tnull as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

[ldk-claude-review-bot]

I've thoroughly reviewed the entire PR diff and the surrounding code. The change extends the v1 data detection in FilesystemStoreV2::new() from checking only the root level to also checking one level down inside namespace directories. This correctly catches v1 layouts like primary/key where v2 expects primary/secondary_ns_dir/key.

I verified:

• The v2 layout always places files at the third level (primary_ns_dir/secondary_ns_dir/key), so the two-level check cannot produce false positives on valid v2 data.
• Temporary .tmp and .trash files from writes are created alongside the destination file (third level), so they won't trigger false positives.
• The [empty] sentinel directory (used for empty namespaces in v2) is properly handled — it's a directory, so it's traversed but not flagged.
• The ambiguous case (v1 with non-empty primary AND secondary) is correctly acknowledged as indistinguishable from v2 data and is not rejected.
• Tests cover all relevant cases: files at root, files one level down, empty directories, nested namespace directories, and valid v2 data with written keys.

No issues found.

[ldk-reviews-bot]

👋 The first review has been submitted!

Do you think this PR is ready for a second reviewer? If so, click here to assign a second reviewer.

[TheBlueMatt]

Heh, I figured this was deliberate to avoid the overhead, but I guess its nice to check.