[ntuple] Add kTypeS3 locator type for S3 storage backend

[JasMehta08]

This Pull Request

(Is a part of the GSoC 2026 project S3 Backend for RNTuple.)

The S3 backend supports two addressing modes: Mode A packs multiple pages into a single S3 object and uses a new locator type to encode both the object ID and byte offset, while Mode B stores one page per object and reuses the existing Object64 locator. This PR adds the on-disk format support needed for Mode A.

This PR:

• Adds kTypeS3 = 0x03 to the ELocatorType enum in RNTupleTypes.hxx
• Wires the new type code into SerializeLocator() and DeserializeLocator() in RNTupleSerialize.cxx, reusing the existing SerializeLocatorPayloadObject64 / DeserializeLocatorPayloadObject64
• Adds round-trip serialization tests for the new locator type, verifying that the 64-bit payload passes the serialize-deserialize cycle with the correct type code

The 64-bit payload is interpreted differently depending on the locator type: 0x03 (Mode A) bit-splices upper 32 bits as object ID and lower 32 bits as byte offset, while 0x02 (Mode B) uses the full value as a plain object ID. The interpretation is handled by the backend, not the serializer. The serializer just stores and retrieves the raw 64-bit value.

Checklist

• Tested changes locally
• Locator round-trip tests pass for both kTypeS3 (0x03) and kTypeDAOS (0x02), locally.

[github-actions]

Test Results

    21 files      21 suites   3d 10h 4m 18s ⏱️
 3 857 tests  3 857 ✅ 0 💤 0 ❌
73 345 runs  73 345 ✅ 0 💤 0 ❌

Results for commit bee8ce4.

♻️ This comment has been updated with latest results.

[JasMehta08]

@jblomer
I have rebased the commits into one and also changed the kTestLocatorType to 5, as suggested. Please have a look and let me know if i should change something.

Thanks!

[jblomer]

In principle looks good to me.

I wonder, however, if it would be better to encode the two modes in two different locator types. We would then consolidate the DAOS with the mode B locator (e.g. kTypeObject64) and create another locator type for mode A (e.g. kTypeMulti or similar). To be discussed, possibly for a follow-up PR.