Auto-detect dimension names from DataArray, remove bracket-spec API

[drbenvincent]

Closes #11. Dimensions are now read from the xarray DataArray and spread automatically — no bracket syntax needed.

# Before
td.parameter_draws(dt, "beta[groups]", "sigma")
td.compare_draws(dt, "beta[groups]")

# After
td.parameter_draws(dt, "beta", "sigma")
td.compare_draws(dt, "beta")

What changed

• Removed _parse_var_spec() (~35 lines of regex + dimension validation)
• Removed import re, from typing import Tuple, List
• parameter_draws: accepts bare *var_names — auto-detects non-sample dims from DataArray.dims
• compare_draws: follows via delegation, renamed *var_specs → *var_names
• prediction_draws: docstring no longer misleadingly mentions bracket syntax
• Tests: removed test_error_malformed_spec, test_error_dimension_mismatch (both test files); all specs updated to bare names
• Docs: all 4 .qmd files + AGENTS.md updated

Verification

• 49/49 tests pass (93% coverage)
• Smoke test matches the exact example from the issue

[drbenvincent]

Good simplification — auto-detecting dims from the DataArray is cleaner than the bracket-spec, and the implementation is correct. A few issues to address before merging.

[drbenvincent]

Version bump needed — removing the bracket-spec API ("beta[groups]" → "beta") is a breaking change for any user on v0.2.x who calls parameter_draws or compare_draws with the old syntax. Per SemVer this should be 0.3.0, not 0.2.0. At minimum the PR description should note the version bump and pyproject.toml should be updated before merging.

Also worth adding a one-line migration note to the changelog / release notes so users know what to change.

[drbenvincent]

Bumped to 0.3.0 in both __init__.py and pyproject.toml. Commit message flags this as a breaking change. A migration note (bare names, no brackets) belongs in the release notes when the tag is cut.