Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@cpp/src/mip_heuristics/presolve/bounds_update_helpers.cuh`:
- Around line 46-47: The hardcoded acceptance scale (candidate_bound_scale =
f_t{1e-14}) in bounds_update_helpers.cuh makes bound gating inflexible; change
the code to derive this scale from the solver's configured tolerances (or accept
a passed-in relative tolerance) instead of the magic constant: replace the
literal with a parameter or a value read from the solver settings (e.g.,
solver_tolerances.relative_bound_tol or an added function parameter rel_tol) and
use that derived value in the existing comparison that uses
candidate_bound_scale, while keeping the existing variables bound and abs_tol
intact.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@cpp/src/mip_heuristics/presolve/bounds_presolve.cu`:
- Around line 121-122: The division computing upd.candidate_bound_scale
(pb.tolerances.absolute_tolerance / context.settings.semi_continuous_big_m) must
be guarded against zero or near-zero denominators; modify the code that sets
upd.candidate_bound_scale to first check context.settings.semi_continuous_big_m
> epsilon (use a small constant like 1e-12 or an existing solver epsilon if
available) and if not, set candidate_bound_scale to a safe clamped value (e.g.
absolute_tolerance / epsilon or a predefined max_scale) so the acceptance
predicate does not get disabled by Inf/huge values; update any related comments
and consider validating semi_continuous_big_m at solver initialization as an
alternative.

In `@cpp/src/mip_heuristics/presolve/multi_probe.cu`:
- Around line 144-147: The division for upd_0.candidate_bound_scale and
upd_1.candidate_bound_scale uses context.settings.semi_continuous_big_m without
guarding against zero/near-zero; update the assignment logic (where
candidate_bound_scale is set for upd_0 and upd_1) to first check that
context.settings.semi_continuous_big_m > epsilon (choose a small epsilon, e.g.
1e-12) and only perform pb.tolerances.absolute_tolerance / semi_continuous_big_m
when that condition holds, otherwise set candidate_bound_scale to a safe
fallback or clamp the computed value into a reasonable range (max_scale) to
avoid infinity/huge values that break the acceptance predicate; reference the
variables upd_0, upd_1, candidate_bound_scale, pb.tolerances.absolute_tolerance,
and context.settings.semi_continuous_big_m when making the change.

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@cpp/src/mip_heuristics/solve.cu`:
- Around line 339-346: The validation currently allows
settings.tolerances.absolute_tolerance == 0 which disables the huge-bound guard;
change the cuopt_expects condition in the existing validation call to require
absolute_tolerance > f_t(0) (strictly positive) instead of >= f_t(0), and update
the error message text in the same cuopt_expects invocation to state that
mip_absolute_tolerance must be strictly positive (finite and > 0) while keeping
the existing checks for settings.semi_continuous_big_m > mip_absolute_tolerance;
this affects the cuopt_expects invocation that references
settings.tolerances.absolute_tolerance and settings.semi_continuous_big_m.