Modeling Algorithms - fix unbounded self-interference: O(1) tangent-zone point access + a checkpointed breaker

[gsdali]

Summary

Fixes #1385: BOPAlgo_ArgumentAnalyzer's self-interference check (SelfInterMode()) can run effectively unbounded on a degenerate self-intersecting surface.

1. Intf_Interference::Insert called Intf_TangentZone::GetPoint(Index) repeatedly inside a nested comparison loop. GetPoint is O(n) per call (the backing NCollection_Sequence has no O(1) indexed access), so every comparison paid that cost again -- measured ~80% of runtime in NCollection_BaseSequence::Find on the artifact in BOPAlgo_ArgumentAnalyzer self-interference: unbounded runtime (no checkpoint) + O(n) tangent-zone lookup #1385. Intf_TangentZone::Points() now caches a true random-access array per zone (invalidated on any mutation): same comparisons, same result, O(1) lookup.

2. The self-interference phase never polled its cooperative progress indicator below BOPAlgo_CheckerSI::CheckFaceSelfIntersection, so a caller's timeout could only fire between whole-face checks, not within one -- 619s CPU against a 30s deadline on the artifact in BOPAlgo_ArgumentAnalyzer self-interference: unbounded runtime (no checkpoint) + O(n) tangent-zone lookup #1385, never returning. Intf_Interference::SetBreaker (RAII-scoped via Intf_InterferenceBreakerScope) lets Insert() poll every 256 calls and abort by throwing Standard_Failure. Wired up in BOPAlgo_CheckerSI's self-intersect functor, only when execution is guaranteed single-threaded -- an exception thrown from a worker thread of OSD_Parallel::For's parallel path would be unsafe (std::terminate).

Verification

On the artifact linked in #1385: with both fixes, a 0.5s deadline returns in 0.547s and a 30s deadline returns in 30.1s (vs. 619s+/never on stock), with correct HasFaulty() results throughout (the 0.5s case correctly reports "not yet determined" rather than a false negative, since the breaker trips before any fault is recorded).

Locally (minimal-module RelWithDebInfo build, FoundationClasses+ModelingData+ModelingAlgorithms only):

• New GTests (Intf_TangentZone_Test.cxx, Intf_Interference_Test.cxx): 8/8 pass.
• Related existing suites (Intf_*, IntPatch_*, IntTools_*, BOPAlgo_*, BRepAlgoAPI_*): 76/76 pass.
• Full OpenCascadeGTest: 7241/7249 pass. The 8 failures are all BRepMesh_DiscretAlgoFactoryTest (mesh discretization plugin self-registration), an artifact of this being a minimal-module local build -- unrelated to this change and not touched by it.

Test plan

• clang-format --style=file applied to all changed files.
• New GTests added and passing locally.
• No regression on related existing GTests.
• CI (opening as draft to check style/build/test results first, per CONTRIBUTING.md).