Render fixed Markdown reports and shared completion counts

internal/core/reporting/testdata/report.md

@@ -0,0 +1,43 @@
+# Audita report: run-123
+
+Status: completed
+
+Corrections applied: 1  
+Work units: 4 total; 2 completed; 1 interrupted; 1 not started  
+Errors: 1 work units; 0 run errors  
+Recoveries: 3 of 4 model operations  
+Proposal responses: 1; with filtering: 1
+
+Transcript input: in\|put.json  
+Glossary input: glossary\`[REDACTED]\`.yaml  
+Transcript output: out<br>file.json
+
+## Work-unit errors
+
+| Chunk | grammar_1 | grammar_2 |
+| --- | ---: | ---: |
+| Chunk 1 (segments 1–2) | 1 | 0* |
+| Chunk 2 (segments 3–4) | 0 | — |
+
+## Recoveries
+
+| Chunk | grammar_1 | grammar_2 |
+| --- | ---: | ---: |
+| Chunk 1 (segments 1–2) | 1 | 2* |
+| Chunk 2 (segments 3–4) | 0 | — |
+
+Legend: `—` = not started; `n*` = interrupted with n recorded errors or recoveries. Completed error cells are 0 or 1.
+
+### Error reasons
+
+- `transport_failed`: 1
+
+### Recovery reasons
+
+- `output_repaired`: 2
+- `proposals_filtered`: 1
+- `service_recovered`: 1
+
+A recovery can have multiple reasons; reason counts need not sum to the recovery total.
+
+Details: `errors.yml` and `recoveries.yml`.