Prevent huge timeout literals from stalling reliability scans

[alxxjohn]

Motivation

• Avoid unbounded math/big.Int parsing of repository-controlled integer literals used to detect zero http.Client.Timeout, which can be maliciously large and cause CPU denial-of-service during scanning.
• Preserve the reliability check semantics (detect zero timeouts) while eliminating the expensive arbitrary-precision conversion on untrusted source text.

Description

• Replace the big.Int parse in isZeroDuration with a lexical check via a new helper isZeroIntegerLiteral, and remove the math/big dependency from internal/codeguard/checks/reliability/reliability_go_helpers.go.
• isZeroIntegerLiteral handles Go integer literal formats (decimal, binary 0b, octal 0o, hex 0x, and underscores) by scanning characters for any nonzero digit and returns true only for literals that are lexically zero.
• Add internal/codeguard/checks/reliability/reliability_go_helpers_test.go with unit tests covering zero/nonzero literals across bases and a regression test exercising a multi-million-digit untrusted literal to ensure the function completes quickly.
• Apply formatting changes with gofmt to the modified files.

Testing

• Ran go test ./internal/codeguard/checks/reliability and the package tests passed.
• Ran go test ./tests/checks and the integration tests passed.
• Ran go test ./... and the full test suite completed successfully.

---

Codex Task