fix: improve scanner precision for Go repositories

[alxxjohn]

Summary

• keep baseline suppressions stable when findings move to split files by adding path-insensitive content fingerprints
• reduce Go scanner false positives for standard result/error returns, local row variables, and Postgres repository/adapter boundaries
• tighten boolean predicate naming to boolean-returning declarations and add common predicate vocabulary

Tests

• szr proxy go test ./tests/checks -run 'TestBaselineSuppressesFindingAfterMoveToSplitFile|TestFunctionReturnContractAllowsStandardGoRepositoryResults|TestBooleanPredicateNamingOnlyRunsOnBooleanReturns|TestBooleanPredicateNamingAcceptsConventionalPredicateVocabulary|TestGoLocalRowAssignmentsDoNotLookLikeMutableGlobals|TestPostgresRepositoryAllowsPgxQueriesAndRowsAffectedResults|TestPostgresStorageAdapterAllowsPgxQueriesOutsideRepositoryFilename'
• szr proxy go test ./internal/codeguard/...
• szr proxy go test <all packages except github.com/devr-tools/codeguard/tests/checks>

Note: full ./tests/checks and go test ./... stayed silent for several minutes in this environment, so the broad verification excludes that long-running package while the new focused tests cover the changed scanner behavior.