[r3wretrhy]

Problem:

• expect(a == _d(1.e19)) hangs (The _d() function seems to cause an infinite loop for large numbers. #711). To pick the default precision, the value constructor calls math::den_size, which multiplies by 10 and casts to unsigned long long until there's no fraction left. Past ~1.8e19 that cast is undefined and on GCC the remainder never hits zero, the loop just keeps going.
• 1.e18 works because 1.e19 still fits. _f has the same issue (1.e20f).

Solution:

• Break out once the value is too big for T. A number that big has no fractional digits anyway. Written as not(value < limit) so inf/NaN stop as well.
• Added a few big values to the _d/_f checks in test/ut/ut.cpp.

Issue: #711

Tested on Linux with GCC 14.2 and Clang 19. With just the test change ut_test hangs (killed it after 20s). With the fix everything builds, ctest passes and the post-build run of examples/benchmarks passes. Also tried the snippet from the issue.

Reviewers:
@