TextValidator/SdkAnalyzer: name offending label ids and summarise missing labels

[rousso]

What

Two related changes that make missing-label problems actionable from the analyzer output.

1. Name the offending identifier in the message. TextValidator now reports which label identifier it found inside a label's text, e.g. Label in RO contains label identifier(s): expression|name|906.

2. End-of-run missing-labels summary. Missing-label identifiers are carried structurally on ValidationResult and aggregated into a deduplicated report at the end of a run, split into two sections:

• Labels found missing from the export — referenced by the SDK but not present (the drools Referenced label … does not exist rules).
• Labels assumed missing during export — identifier left in label text by the exporter when it could not resolve the label (TextValidator).

Each identifier is listed once with its reference count.

3. --verbose flag. Without it, the thousands of individual missing-label errors are suppressed in favour of the summary, and a hint points to --verbose for the per-occurrence detail. The error count and exit code are unchanged either way.

Notes

• The two checks are disjoint in practice (expression ids via the text fallback; business-entity/rule-text ids via structural references), so the found/assumed split is clean.
• Tests: existing tedefo-3301 scenarios and the full suite (177) pass; TextValidatorTest extended to assert the named ids and the ASSUMED kind.