* pdf inspection: count pending text clips and drop page content after reading
- hayro-interpret: state_bytes counts the pending text-clip outlines a
saved state copies, so MAX_SAVED_STATE_BYTES bounds them too.
- plate-test structure: read each page through a ContentReader into a
buffer dropped after the page, instead of page.operations(), which keeps
every page's decoded content for as long as the document.
* address review: cap page content read at the 256 MiB decode limit
* fix the saved-state test: start the clip outline with a MoveTo
Dylan M. Taylor
7f0f352d21