Collapse map-data onto a single writer API

[amc-corey-cox]

Closes #302.

map-data used two different writer stacks depending on whether -O was passed. The -O path built a MultiStreamWriter; the single-output path used the function-based writers and a hand-rolled header-rewrite block that only worked because tsv_stream/csv_stream stash their headers on the function object.

This routes both through MultiStreamWriter — the single-output case is just the multi-output case with no additional outputs. MultiStreamWriter already handles the tabular header rewrite generically, so the inline rewrite and the get_stream_writer / rewrite_header_and_pad / nullcontext / os imports drop out.

The function-based API (yaml_stream, json_stream, jsonl_stream, tsv_stream, csv_stream, get_stream_writer, and the TabularStreamWriter.stream alias) is now unused internally. Per the decision on #302 it is deprecated, not deleted: the names stay exported from linkml_map.writers, now emitting a DeprecationWarning that matches the repo's existing "removed in a future version" wording. The raw line reduction lands later when the deprecation cycle completes.

Testing

• New CLI-level test for tabular output over heterogeneous records (records with different keys) — a path that previously had no coverage. It asserts the resulting table order-independently, since output column order varies by environment.
• TestDeprecatedFunctionAPI asserts each deprecated name warns; the redundant function-API tests are marked to ignore the warning so the suite stays quiet.
• Full suite: 1078 passed, 4 skipped, 1 xfailed. Lint, format, and doctests clean.

Note for reviewers

While writing the CLI test I hit a pre-existing quirk unrelated to this change: tabular column order varies by environment (subprocess CLI emits insertion order, in-process CliRunner under pytest emits sorted order). The object keys and values are identical either way — only column ordering differs, which is cosmetic. The new test is order-independent because of it. Possibly worth its own issue.

[Copilot]

Pull request overview

Unifies the map-data CLI output pipeline by routing both single-output and -O/--additional-output cases through the class-based MultiStreamWriter, while keeping the legacy function-based writer API available but deprecated (per #302). This reduces duplicated writer logic and centralizes tabular header-rewrite behavior.

Changes:

• Refactor map-data streaming output to always use MultiStreamWriter (single output becomes the degenerate multi-output case).
• Deprecate function-based writer entry points (*_stream, get_stream_writer, TabularStreamWriter.stream) by emitting DeprecationWarning, and adjust tests to either assert or silence warnings as appropriate.
• Add a CLI-level TSV regression test covering heterogeneous records (late-discovered tabular columns) with order-independent assertions.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
src/linkml_map/cli/cli.py	Removes the separate single-output writer path and routes all streaming output through MultiStreamWriter.
src/linkml_map/writers/output_streams.py	Adds a shared deprecation-warning helper and deprecates legacy function API + TabularStreamWriter.stream.
tests/test_cli/test_cli_tabular.py	Adds an integration test ensuring TSV output rewrites/pads correctly when columns are discovered late.
tests/test_writers/test_output_streams.py	Updates tests to handle deprecations (silence in legacy tests, assert warnings in deprecation tests).

Comments suppressed due to low confidence (1)

src/linkml_map/writers/output_streams.py:469

• csv_stream() retains key_name for API compatibility but no longer uses it after switching to writer.process(chunks). To prevent Ruff unused-argument warnings (ARG001), mark the parameter with # noqa: ARG001 (or rename to _key_name in a future breaking release).

    """
    Convert chunks of objects to CSV format.

    .. deprecated::
        Use ``make_stream_writer(OutputFormat.CSV)`` and drive it via