get_page_image: add bbox crop parameter (normalized [x0,y0,x1,y1])

Render only a cropped region of a PDF page via PyMuPDF clip rect.
Coordinates are normalized 0-1 relative to the page, matching the bbox
returned by zotero_get_page_layout. Keeps rendered figure images small
so they can be embedded inline as base64 data URIs.

src/pdf_mcp/server.py

@@ -207,8 +207,9 @@ def get_page_image(
     page: int = 1,
     dpi: int = 150,
     output: Literal["base64", "file"] = "base64",
+    bbox: list[float] | None = None,
 ):
-    """Render a single PDF page as a PNG image.
+    """Render a single PDF page (or a cropped region) as a PNG image.
 
     Args:
         filename: Path to a PDF file.
@@ -216,10 +217,27 @@ def get_page_image(
         dpi: Image resolution. Default 150 (good balance of readability and size).
         output: "base64" returns the image inline as MCP image content.
                 "file" writes to a temp file and returns the path.
+        bbox: Optional crop region as [x0, y0, x1, y1] in normalized
+              coordinates (0-1 relative to the page). Renders only that
+              region, which keeps the image small — ideal for a figure or
+              table. Use zotero_get_page_layout on the Zotero backend to
+              get the bbox of a figure. If omitted, renders the full page.
     """
     doc = fitz.open(filename)
     pdf_page = doc[page - 1]
-    pixmap = pdf_page.get_pixmap(dpi=dpi)
+    if bbox:
+        x0, y0, x1, y1 = bbox
+        # Normalized (0-1) coords -> points
+        pr = pdf_page.rect
+        clip = fitz.Rect(
+            pr.x0 + x0 * pr.width,
+            pr.y0 + y0 * pr.height,
+            pr.x0 + x1 * pr.width,
+            pr.y0 + y1 * pr.height,
+        )
+        pixmap = pdf_page.get_pixmap(dpi=dpi, clip=clip)
+    else:
+        pixmap = pdf_page.get_pixmap(dpi=dpi)
     png_bytes = pixmap.tobytes("png")
     doc.close()