[Bug] S3 FileIO fails with Wrong FS when one FileIO reads both s3:// and s3a:// paths of the same bucket

[thswlsqls]

Search before asking

• I searched in the issues and found nothing similar.

Paimon version
master @ 9e306a4

Compute Engine
Engine-agnostic (S3 filesystem plugin); surfaced by Hive clone.

Minimal reproduce step

1. Create a HiveCatalog with warehouse s3://bucket/wh. HiveCatalog.createHiveCatalog builds one FileIO and checkOrMkdirs(warehouse) caches the s3 FileSystem.
2. Clone a Hive table whose HMS location is s3a://bucket/.... CloneFilesFunction and HiveCloneUtils.listFiles read it through the same hiveCatalog.fileIO().
3. Fails with IllegalArgumentException: Wrong FS s3a://bucket/... -expected s3://bucket.

Cause: HadoopCompliantFileIO.getFileSystem() (paimon-s3-impl) keys fsMap by authority only, while each S3AFileSystem is bound to its scheme. Core HadoopFileIO already keys by Pair.of(scheme, authority) (#2504).

What doesn't meet your expectations?
One S3 FileIO should serve both schemes, since the S3 plugin also serves s3a:// (#9814). Actual: the first cached scheme wins.

Anything else?

• Also reachable via data-file.external-paths with another scheme, only when resolving-file-io.enabled=false; ResolvingFileIO already keys by scheme and authority.
• Cross-scheme rename (e.g. Hive migrate, s3a to s3) still fails; out of scope.

Are you willing to submit a PR?

• I'm willing to submit a PR!