[Bug] Avoid altering lake catalog for tables that never enabled datalake

[MaheshS08]

Purpose

Linked issue: close #3297

The preAlterTableProperties method was calling lakeCatalog.alterTable()
for every table alter operation whenever a lake catalog was configured at the
cluster level, even for plain Fluss tables that never had table.datalake.enabled
set. This resulted in unnecessary calls to the lake catalog, which would throw
TableNotExistException (silently swallowed) for tables unknown to the lake catalog.

Brief change log

• In MetadataManager#preAlterTableProperties, tightened the guard condition
  from if (lakeCatalog != null) to also check that the current table descriptor
  contains the table.datalake.enabled key (regardless of its value).
• This ensures lakeCatalog.alterTable() is only called for tables that have
  ever opted into datalake, while preserving the existing sync behavior for
  tables that were previously lake-enabled and later disabled.

Tests

• No Tests required

API and Format

No API or storage format changes. This is a behavioral fix in the coordinator's
metadata management logic only.

Documentation

No new feature introduced. No documentation changes required.

[MaheshS08]

@luoyuxia , could you please review this PR?

[luoyuxia]

@luoyuxia , could you please review this PR?

But ci fails

[Prajwal-banakar]

Hi @MaheshS08 Looks like the CI failure is from Spotless formatting. You may have missed running mvn spotless:apply before pushing the changes.

Also, please go through the contributing guidelines once most of these checks are covered there and it helps avoid CI failures

[MaheshS08]

@Prajwal-banakar thanks for the head's up. I have re-pushed to the branch with necessary changes.

@luoyuxia , please have a look at this PR, now the checks have been passed.