Add toDouble for the FastInt family

[meshinspector-agent]

First of the three increments that restore the shadow filter's double pre-filters. They were dropped in the port because the FastInt family has no conversion to double — on MSVC FastInt128 is std::_Signed128, which has no such operator at all — so every shadow test is currently exact.

What changed

• toDouble( FastInt128 ) and toDouble( const FastInt<nBits> & ), both over one shared detail::doubleFromWords.
• Correctly rounded: exact below 2^53, relative error below 2^-53 above it, ±inf past DBL_MAX. The bound is documented at the function, since a pre-filter that rejects a case in double is only safe against a stated conversion error.
• Deliberately not the built-in __int128_t → double on gcc/clang: the same code on every platform, so the filter's numbers cannot diverge between MSVC and the rest.
• Class comments in MRFastInt.h / MRFastInt128.h no longer claim the conversion is missing.

Correct rounding comes from a single std::uint64_t → double (itself correctly rounded) of the top 64 significant bits, with everything below folded into a sticky lowest bit — so a tie is told from a value just above it without a second rounding.

No caller yet: increments 2 and 3 consume it.

Verification

• Built MRTest in Release with g++-12 in the ubuntu22 image, mirroring build-test-ubuntu-x64.yml (prebuilt thirdparty linked, Python bindings off) — clean.
• Full MRTest: 354/354 pass, including MRAlphaShapeTests and the InSphere/SoS predicate tests.
• 7 new tests in MRFastIntTests.cpp, against that file's existing RefInt oracle rather than any code under test:
  • exactness below 2^53 at every width, and +0 for zero;
  • per value: the result is nearest — neither std::nextafter neighbour is closer — and within the documented 2^-53;
  • the tie cases at 2^53 (round-to-even both ways) and the sticky-bit tie at 2^64 + 2^11, which only a correct tail test gets right;
  • sign symmetry, and the same value at different widths converting to the same double;
  • ±inf past DBL_MAX — reachable only through a product, since a 1024-bit value maxes at 2^1023 < DBL_MAX, so FastInt<2048> and max1024 * max1024 cover it.
• Not exercised here: MSVC's std::_Signed128 path (no Windows box on this runner) — sharing one word-based implementation is precisely what makes that path the same code; CI covers it. No UI surface and nothing to drive in the viewer: a header-only numeric addition with no caller until the next increment, so behaviour is unchanged at runtime.
• No timings from this box — it is not a GitHub-hosted runner.

CI label: none. build-test-distribute.yml is gated on pull_request: branches: [ master ], so a PR based on alpha-shape-triangle-shadow gets no checks by design; the numbers come from the branch run after the squash-merge.

[meshinspector-agent]

Superseded by #6594 — same commit rebased onto master, so it actually gets checks. Force-push to agent/* is blocked by a repo rule, hence a new branch rather than a rebase in place. Branch kept alive because #6593 is based on it.