feat: model TM1 12.6.1 Arrow/Parquet/Flight TI datasources

[skido-cw]

Summary

Adds first-class modelling for the three new TurboIntegrator input datasource types introduced in TM1 v12, build 12.6.1 — Apache Arrow IPC/Feather (ARROW), Apache Parquet (PARQUET), and Apache Arrow Flight (FLIGHT). This is pure object modelling of the Process DataSource block; no transport coupling and no new service methods. Previously these types fell through to an empty "DataSource": {}.

Changes

• Process: four new datasource_flight_* scalars (constructor, from_dict, properties) and two new _construct_body_as_dict branches:
  • ARROW/PARQUET — file name or http(s) URL via dataSourceNameForServer/dataSourceNameForClient, plus optional jsonRootPointer / jsonVariableMapping for nested (struct/list/map) columns; no ascii* keys.
  • FLIGHT — flightLocation / flightDescriptorType / flightDescriptor / flightAuth.
  • Both branches match Type case-insensitively (see "Verified" below).
• ProcessService.get / get_all: select the four flight fields.
• datasource_type stays a free string (no Type whitelist) and no version gating is added — ProcessDataSource is an OData open type whose Type is validated by the engine at execute time.

Verified against a live PA 12.6.1 server

End-to-end create → read-back confirmed two behaviors worth flagging:

• The Arrow Flight wire fields are flightLocation / flightDescriptorType / flightDescriptor / flightAuth (a dataSourceFlight* payload is silently dropped).
• The server canonicalizes Type to title-case on read-back (Arrow / Parquet / Flight), which is why the body branches match case-insensitively — otherwise from_dict(get(...)) would round-trip to an empty DataSource.

Testing

• Offline unit tests (no server) for the body shapes, from_dict ⇄ body_as_dict round-trips, and the title-case Type round-trip — 17 tests, all pass.
• Optional live round-trip test gated to TM1 ≥ 12.6.1 (@skip_if_version_lower_than), self-cleaning — passes against a live 12.6.1 server.
• black --check . and ruff check . pass.

Backward compatibility

No changes to the existing datasource branches (ASCII / None / ODBC / TM1CubeView / TM1DimensionSubset / JSON); the public API stays byte-compatible for every existing datasource type.

🤖 Generated with Claude Code

[skido-cw]

Superseded by #1431, which contains this datasource-modelling commit plus the new write_dataframe(use_parquet=/use_arrow=) columnar blob-write path. Closing in favor of the combined PR.