Implement `failIfAlreadyExists` in S3 repositories

[DaveCTurner]

Several write methods on BlobContainer accept a failIfAlreadyExists parameter to prevent Elasticsearch from overwriting an existing blob. In the past, the S3 API had no way to express this constraint so S3BlobContainer simply ignores it and hopes for the best. However since late 2024 AWS S3 now supports conditional writes, as does MinIO, so we should make use of this new feature in Elasticsearch for extra repo-corruption protection.

In terms of testing, we should verify that we are really doing conditional writes by adding a test to S3BlobStoreRepositoryTests, adjusting S3HttpHandler as required, and therefore adding to S3HttpHandlerTests too. Since all the other repository types support conditional writes, with this change we can safely add a test for conditional writes in org.elasticsearch.repositories.AbstractThirdPartyRepositoryTestCase. Ideally we would also extend RepositoryAnalyzeAction to verify conditional writes work correctly in real clusters, but we could leave that for a follow-up PR if desired.

[elasticsearchmachine]

Pinging @elastic/es-distributed-coordination (Team:Distributed Coordination)

[DaveCTurner]

With #133030 now merged, I'm moving the remaining task from this issue into its own ticket at #134632.