f2f64fb1

By: ValentinHirschi <valentin.hirschi@gmail.com>

bind conditional cut and side sampling in native frames

Resolve cut/side/active blocks once through the Symbolica channel model and one
qualified geometry registry. Prepare each conditional map from its actual raw
prerequisites, solve the native host LU root, and retain the complete external
shift in the physical-to-raw affine transformation. Use the same supplied-point
preparation for foreign inverse densities and route each complete native map
to the master frame once. Preserve the complete physical raised-cut sum/events.

Resolve ordered complete parents through existing generated routing. Cache
numerical frozen-binding failures alongside native successes so the existing
source retry can rescue them; structural errors and unusable epochs still fail.
Remove the unused complete prepared-sample handoff. Correct the bridge Gaussian
raw second moment and retain typed numerical errors through profile compilation.

Validation:
- 141 core/API tests passed, including shifted saved-state acceptance, physical
  summed/explicit channel equality, actual nine-dimensional conditional cuts,
  boosted alternate frames, LU-h composition, foreign inverses, full Cartesian
  Jacobians, native range rescue, and actual frozen-kite warmup rescue.
- All 11 test_differential tests passed.
- Core/API test checking with python_api, differential test checking, formatting
  and clippy pass. The map-enum size warning is unchanged from the prior commit.
- Existing runtime-cache payloads still encode zero bytes; no saved physical
  schema change. Internal frame-map and compile APIs are migrated consistently.

The generated conditional fixture keeps its 8192 draws per channel/frame and
uses the existing five-minute acceptance timeout. Broad regressions update only
obsolete diagnostics, retaining dependency and unsupported-primitive checks.

GL638 direct-H validation is next; this milestone does not establish GL638
variance improvement or finish joint/star maps, automatic discovery, or the
all-orientation/full-UV goal. The plan and research ledger retain those limits.

User prompts:

Continue as planned, but re-read the plan, and establish how to best orchestrate progress on both front of amplitude surface advanced sampling, but also the overall goal of cross-section advanced sampling, that must eventually apply to GL638 and demonstrate improvement there.

(continue as planned, but remember that for being able to quickly generate states for testing, specifying ot one orientation may help. But ultimately, things must be tested end-to-end on all orientations)

(oh and I forgot, you should commit+push regularly at good milestones in a new feature branch `advanced_sampling`, which does not exist yet )

Add to do all git commit and push from the user ValentinHirschi with email valentin.hirschi@gmail.com
Also explicitly add to the plan that this is a hard task and you should consult the existing research you did already often (specifiy relevant markdown) and complement it when needed and use many subagents to do implementation slices, code audit, tests and research. As the main agents you should mostly act as an orchestrator, so as to not pollute your context.

Timeline