Fixes #2518 .
Breaking fix for users who have adapted to the inverted behavior when supplying min_t

Fixes the inverted behavior of min_t argument in AABB::intersect_ray.

The min_t value passed in is used as t1 when calling ray_box_intersect which sets the upper limit of t range considered for intersections. That mixup is part of a larger issue with a mismatch in intended behavior and usage of min_t between the public intersect_ray and private intersect_ray_opt called recursively to do the search and return the nearest hit.

L1686 in AABB.cpp is attempting to back out the nearest hit value with comparisons to min_t. That's consistent with the defined behavior for that specific intersect_ray_opt overload, but clashes with the expected behavior of min_t in the public method intersect_ray.

This patch introduces a max_t parameter to the intersect_ray_opt method. Internally BVH culling is now done with comparisons to max_t and the user input of min_t is properly passed as the lower limit, t0, of the ray_box_intersect call. The PR also includes a separate fix for a hardcoded cast to float in ray_triangle_intersect.

The existing unit test AABB: intersect_tet is updated to pass min_t = 0 instead of min_t = std::numeric_limits<double>::infinity() which follows the old, inverted behavior. A new unit test (AABB: intersect_cube_min_t) is added following the example given in the bug report (#2518)

Checklist

• All changes meet libigl style-guidelines.
• Adds new .cpp file.
• Adds corresponding unit test.
• This is a minor change.