Show progress bar for full-page OCR

[iRonin]

Closes #511.

Show a progress bar for full-page OCR

What

Add a tqdm progress bar to chat_completions_batch (the dispatch used by
full-page OCR / surya_ocr), gated by the existing settings.DISABLE_TQDM.

Why

The recognition / full-page path was the only batched path with no progress
indication — detection, layout and the OCR-error model already use tqdm. On a
multi-page PDF surya_ocr appeared to hang because nothing is printed until the
whole job finishes. In full-page mode the batch is one item per page, so
total=len(batch) is the page count and each completed item is a finished page.

Change

-    with ThreadPoolExecutor(max_workers=max_workers) as executor:
-        return list(executor.map(_process, batch))
+    with ThreadPoolExecutor(max_workers=max_workers) as executor:
+        return list(
+            tqdm(
+                executor.map(_process, batch),
+                total=len(batch),
+                desc="Recognizing Text",
+                disable=settings.DISABLE_TQDM,
+            )
+        )

(plus the two imports: from surya.settings import settings, from tqdm import tqdm)

Notes

• Respects DISABLE_TQDM, so users who set it keep silent output.
• executor.map preserves the existing result ordering; tqdm advances as the
  in-order results are yielded. (Could use as_completed for finer-grained
  updates, but that would require re-sorting outputs by metadata and is a
  larger change — happy to do that instead if preferred.)
• chat_completions_batch is shared with the per-block fallback path, so the
  bar counts requests generically; in the dominant full-page path that is pages.

Tested

surya-ocr 0.20.0, llamacpp backend, Apple Silicon (MPS), 6-page scanned PDF,
SURYA_INFERENCE_PARALLEL=4:

Recognizing Text:   0%|          | 0/6 [00:00<?, ?it/s]
Recognizing Text:  17%|█▋        | 1/6 [01:17<06:28, 77.77s/it]
Recognizing Text: 100%|██████████| 6/6 [02:12<00:00, 22.01s/it]

Output (results.json / markdown) unchanged.

[github-actions]

CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅

[iRonin]

I have read the CLA Document and I hereby sign the CLA