[core] Fix drop table when table directory is missing on disk

[ArnavBalyan]

Purpose

• Drop fails when table exists in metastore but not available on disk
• Ensure we can can check with the hms/jdbc catalog for the entry during deletion
• Also delete the metadata when the table drop is requested but the table does not exist on disk.
• This avoids leaking metadata, and manual risky operations on the metastore to do cleanup.
• Fixes user reported issue [Bug] When using the Hive catalog, after deleting the table directory, it is no longer possible to drop the table #4853

Tests

• UT

[JingsongLi]

I think this catch block is too broad for a destructive operation. dropTable now catches TableNotExistException | RuntimeException while collecting external paths, and then proceeds to dropTableImpl(...) anyway.

That covers the intended "table directory/schema files are already missing" case, but it also swallows real failures such as schema parse errors, branch listing failures, permission errors, or transient filesystem/runtime errors from schemaManager().listAll(), branchManager().branches(), or getSchemaExternalPaths(...). In those cases the code may remove the catalog/table metadata while silently losing the external path information that was supposed to be protected/cleaned.

Could we narrow this to the specific missing-table/missing-path condition this PR wants to tolerate, and let other runtime failures abort the drop?