Core: CatalogTests: invalidate cache after renaming metadata file

[adutra]

This fix prevents test failures when catalogs have a table cache. This issue may arise e.g. with RESTCatalog since the recent introduction of RESTTableCache in 72d5fd6.

[pvary]

@adutra: Is this flaky? Or how can we not detect this issue during the tests there?

CC: @gaborkaszab

[adutra]

@pvary I don't think it's flaky, it's just that the issue will manifest only with a RESTCatalog that has caching by ETag enabled, which is not the case in TestRESTCatalog (but is the case in the Apache Polaris repo, that's where I spotted the issue).

[pvary]

@pvary I don't think it's flaky, it's just that the issue will manifest only with a RESTCatalog that has caching by ETag enabled, which is not the case in TestRESTCatalog (but is the case in the Apache Polaris repo, that's where I spotted the issue).

Shall we add a test which fails without the change? Otherwise what would prevent someone to break this again for Polaris?