PyAutoLens v2026.10.16.1

No direct changes in this release.

Upstream Changes

docs: link each capability-matrix search to its autofit_inference catalogue card (#1696)
test: drop stray dynamic_delta/delta kwargs passed to DynestyStatic (#1695)
refactor: close the run(ctx) framework, new-search template (search-ext A5 PR 4/4) (#1692)
refactor: MultiStart onto run(ctx), split into phases (search-ext A5 PR 3/4) (#1691)
refactor: Dynesty, BlackJAX NUTS, SMC onto run(ctx) (search-ext A5 PR 2/4) (#1690)
refactor: Emcee, Zeus, BFGS/LBFGS onto run(ctx) (search-ext A5 PR 1/4) (#1689)
refactor: A4c family bases out of the MRO (deprecated thin subclasses, per-posterior_kind defaults and plotter) (#1687)
feat: A4b search-level seed (SeedSequence fan-out, conditional identifier walk, legacy table) (#1686)
refactor: A4a decompose NonLinearSearch (TestModeBypass, EP adapter, runtime settings, paths capabilities) (#1685)
fix: A2/A3/A3b post-merge fixes from astra review 05 (search-extensibility) (#1683)
feat: add n_effective to DynestyDynamic; fix truncated runs under a finite iterations_per_full_update (#1665)

perf: reuse one Cholesky factor of F+H for the solve and the log det (#634)
feat: route interferometer numba vs FFT curvature by a cost model (#632)
fix: exclude zero-weight Delaunay mappings from adaptive pixel signals (NumPy/JAX log-det parity) (#631)
perf: interferometer sparse data vector without the dense mapping matrix (#630)
test: pin DelaunayNN routing to the numba likelihood paths (#623 phase 2) (#625)
perf: numba Sibson kernel so DelaunayNN matches Delaunay speed on CPU (#623 phase 1) (#624)