♻️ Unify register-reference parsing between qubit targets and classical conditions

[rturrado]

🤖 AI text below 🤖

Problem statement

Two places in src/common/parsing/CodePreprocessing.cpp parse a register reference of the form name or name[index] with almost identical logic:

• validateTargets does it for qubit targets in gate calls (x q[0], cx q[0], q[1], assert-ent q[0], q[1], ...). It parses the shape (find [, find ], check the bracket, check that the index is digits, convert to size_t) and then validates the result against definedRegisters (existence + bounds).
• parseBitRegisterRef does the first half of the same work for classical bits referenced in an if condition. It was extracted in the refactor for ✨ Support if (c) and if (c[k]) as boolean-only conditions #462 without noticing that the structural-parsing logic already lived inside validateTargets.

The result is duplicated code paths that treat qubits and classical bits differently even though the parsing shape is the same. Any new check (lexical, semantic, or otherwise) has to be added in two places, and the two sides drift.

Surfaced while planning #468.

Depends on #462: this refactor builds on the parseBitRegisterRef helper introduced in the PR that closes #462, so it can only land after that PR is merged.

Proposed solution

Extract the shared shape-parsing logic into a single helper (working name: parseRegisterRef) that returns something like the current BitRegisterRef (a name plus an optional index). Both validateTargets and the classical-condition path call it. validateTargets continues to apply its qubit-specific validation (definedRegisters, bounds) on top of the parsed result; parseBitRegisterRef can either become a thin wrapper or disappear.

Behavior is unchanged; this PR would be a pure refactor.

Out of scope

• Adding new lexical or semantic checks. That is tracked in 🐛 Register-reference validation is incomplete in the parser #468 and will be easier to implement on top of the unified helper.
• Renaming or restructuring beyond the shared helper.

[rturrado]

Handled as part of PR #463: the refactor was small enough that we merged the commit into the branch of the if (c) PR to avoid the overhead of a separate PR/issue for a tiny change.