Why chimera sometimes need scan the whole buffer with an anchored regex ?

Hi,
Chimera sometimes need scan the whole buffer when a regular expression contains greedy pattern, for example, use a very simple pattern `e2*e` to match `a4444444444444444444444444444444444444444444444444444444444a`. vermSearchAligned and vermUnalign in vermicelli_sse.h will read every char, but this seems no necessary, it should failed at the first char, and 2 is not 4 obviously. When pattern changed to anchored "\Ae2*e", scanDoubleFast in noodle engine will read every char too. Both make the performance very poor. Imagine there is millions of 4.
Is this a bug or designed, and is there any way to improve it? Thanks.

Reproduce the greedy-pattern case with `e2*e` and the anchored `\Ae2*e` example against a long buffer of `4` characters. Inspect `vermicelli_sse.h`, especially `vermSearchAligned` and `vermUnalign`, then trace `scanDoubleFast` in the noodle engine. Done means establishing whether the full-buffer scans are intended and identifying a justified performance improvement if they are not.

Written by the indexing model from the issue text.