Context: comment thread at https://chromium-review.googlesource.com/c/chromium/src/+/1265560#message-cc790779746126dd0a9bbf5ffbef3fafb6428d27

The code in this CL assumes that all of the addresses without CFI information are function prologs.

1. Try to make pseudo-CFI for the <0.1% of stuff not covered that is not a prologue. To make the assumptions more solid. This experiment might cause size regression on the CFI tables.

2. Compile time test: disassemble the whole library, collect uncovered offsets, check they are at function beginning and only contain a whitelist of instructions. Figure out if the test can be written on chrome side or breakpad side for unwind rule coverage.