Route add_enforced_linear_constraint through the enforced helper setters

[winklemad]

Fixes #5302

Model.add_enforced_linear_constraint and the BoundedLinearExpression branch of _add_enforced_linear_constraint_to_helper wrote the indicator constraint's name, bounds and terms through the non-enforced helper setters (set_constraint_name / set_constraint_lower_bound / set_constraint_upper_bound / add_terms_to_constraint) while passing the enforced constraint's index.

Enforced constraints live in MPModelProto.general_constraint[i], but those setters address MPModelProto.constraint[i]. The index spaces are disjoint, so the data landed on an unrelated linear constraint — silently destroying it and leaving the indicator constraint empty — or, when no linear constraint existed at that index, mutable_constraint(i) read out of bounds and the process segfaulted (the DCHECK is compiled out in the release wheel).

This change routes every setter in the enforced paths through the set_enforced_* / add_terms_to_enforced_constraint variants, matching the bool branch of the same helper (which was already correct) and the Java and C# ports.

Added two regression tests:

• test_add_enforced_linear_constraint_without_regular_constraint — an enforced constraint on a model with no prior regular constraint now builds a well-formed general_constraint instead of crashing.
• test_add_enforced_linear_constraint_keeps_prior_constraint — a prior regular constraint keeps its own name, bounds and terms, and the enforced constraint carries its own, instead of the two being conflated.

Both fail on main (the first as a segfault, the second on the corrupted proto) and pass with this change.