Log-space Taylor override hook + API-docs reorg

[thisiscam]

Follow-ups after #2 (two commits).

1. Log-space generator_taylor_coefficients override

Adds Copula.log_generator_taylor_coefficients(t, k_max) returning (sign, log|c|) for the normalized Taylor coefficients c_k = ψ^(k)(t)/k!. The Bell density already consumes coefficients as (sign, log|c|) internally, so a family whose series is computed in the log domain (e.g. AMH's Geometric-frailty logsumexp) can feed them straight in instead of exponentiating to a possibly-overflowing float64 only for bell to take the log again. bell._root_assembly now prefers this hook → linear generator_taylor_coefficients → jet.

Verified (AMH, flat copula, θ=0.5): the log override reproduces the linear override's log-likelihood to machine precision across d=8–50, while plain jet drifts at d=30 and is wrong at d=50 — i.e. it matches the paper's AMH precision result. Regression test in tests/test_log_taylor_override.py (fast: 2 passed in ~2s).

2. API-reference reorganization + install fix

• mkdocstrings: hide private/dunder members, Args/Returns as tables, symbol-type headings + TOC, no source dumps, drop merge_init (was leaking CompiledModel's private dataclass fields).
• api.md: restructured into Defining a copula / Compiling & evaluating / Composition registry / Configuration / Advanced (the generator-precision override hooks, grouped + explained — generator_taylor_coefficients recommended; log_generator_kth_derivative flagged legacy single-layer).
• README + docs/index.md: corrected install — pip install (incl. git+…) does not work yet (ignores the [tool.uv] oryx pin); documents uv-from-source; PyPI gated on an upstream oryx release for jax 0.8.

mkdocs build --strict clean; llms.txt generates.