feat(devserver): support remote dataset version and environment selection

[Curtis Galione (cjgalione)]

Resolves https://linear.app/braintrustdata/issue/SDK-95/remote-eval-dataset-version-and-environment-selection-not-preserved

Summary

Adds dataset version and environment selection to Python remote evals, matching the JavaScript SDK behavior.

• Forwards dataset_version and dataset_environment for both named and ID-based dataset references.
• Adds environment selection to init_dataset() and lazily resolves it to a pinned dataset version.
• Preserves _internal_btql and inline dataset behavior.
• Gives an explicit dataset version precedence over an environment selector.
• Includes org_name when resolving dataset environments and omits it when unavailable.
• Adds end-to-end local HTTP coverage for name/ID references, selector precedence, URL encoding, and optional organization query shaping.

JavaScript SDK parity

This follows the JavaScript devserver and dataset-selection semantics:

• both reference forms forward version and environment selectors;
• explicit version wins over environment;
• environment names are URL-encoded;
• environment resolution includes org_name only when set.

Validation

cd py
mise exec -- nox -s test_cli -- -k test_get_dataset
mise exec -- ruff check src/braintrust/devserver/test_dataset.py src/braintrust/logger.py
mise exec -- ruff format --check src/braintrust/devserver/test_dataset.py src/braintrust/logger.py

7 passed; Ruff lint and formatting checks passed.