[gabotechs]

Which issue does this PR close?

Follow-up to #24631 and #24628. No separate issue.

Rationale for this change

Custom DataSource implementations can encode themselves through try_to_proto, but still need a PhysicalExtensionCodec to decode. Applications combining sources from multiple libraries therefore have to compose their codecs.

Extend the session-scoped decoder registry to sources so each source can own its decoder and register it by name, following the extension execution-plan and expression APIs.

What changes are included in this PR?

• Add ExtensionDataSourceFromProto, register_data_source, and extension_data_source_node. Constructors receive a typed PhysicalExtensionDataSourceNode and return Arc<dyn DataSource>.
• Register DataSourceExec as a shared extension-plan decoder. Its payload contains the source name and bytes; the wrapper looks up the source decoder and rebuilds DataSourceExec through the existing plan registry dispatch. No new top-level plan variant is needed. Source names are independent of plan and expression names.
• Once the wrapper is registered, missing source names, unregistered sources, and decoder errors fail directly. If the wrapper itself is unregistered, existing plan-codec fallback applies to the wrapper payload. Built-in scans and existing codec-encoded sources retain their wire formats.

What is the testing strategy for this PR?

Six registry unit tests and fourteen integration tests cover registration, execution after decoding, reading-session state, malformed wrapper and source payloads, codec dispatch boundaries, legacy codec compatibility, shared names across extension kinds, and binary/JSON round trips.

Validation: 541 affected-crate tests and 14 doctests passed; formatting, workspace-wide Clippy (cargo clippy --all-targets --all-features -- -D warnings), rustdoc with warnings denied, protobuf regeneration, and documentation formatting passed.

Are there any user-facing changes?

New source-registry APIs behind the existing proto feature. Built-in source serialization is unchanged. Readers must register source decoders before writers emit the new wrapper payload. Documented in the 56.0.0 upgrade guide.

AI-assisted implementation with Codex.

[gabotechs]

cc @adriangb I want to tackle this one myself in order to get some hands-on exp with the approach here. Still iterating on this, so not looking for a review.