False positive for SCREAMING_SNAKE_CASE with a trailing lowercase s

[mkatychev]

First encountered here:
https://github.com/RustCrypto/formats/blob/e55540ecc2f5950fad5bc380592df5efc6f68105/pem-rfc7468/README.md?plain=1#L43

1  D COULDa
2  D COULDd
3  D COULDs
4  D SHOULDs
5  I LOCIs
6  I OCTOPId
7  L AT_WILLs
8  L WILLs
9  T BROUGHTd
10 T BROUGHTs
11 T MUSTs
12 T WITs

The README.md above produces the output below:

$ typos README.md
error: `COUL` should be `COULD`
  --> README.md:1:6
  |
1 | 1  D COULDa
  |      ^^^^
  |
error: `COUL` should be `COULD`
  --> README.md:2:6
  |
2 | 2  D COULDd
  |      ^^^^
  |
error: `COUL` should be `COULD`
  --> README.md:3:6
  |
3 | 3  D COULDs
  |      ^^^^
  |
error: `SHOUL` should be `SHOULD`, `SHAWL`, `SHOAL`
  --> README.md:4:6
  |
4 | 4  D SHOULDs
  |      ^^^^^
  |
error: `WIL` should be `WILL`, `WELL`
  --> README.md:7:9
  |
7 | 7  L AT_WILLs
  |         ^^^
  |
error: `WIL` should be `WILL`, `WELL`
  --> README.md:8:6
  |
8 | 8  L WILLs
  |      ^^^
  |

Cases 5,6 and 9-12 seem to be handled correctly.

It seems that a lowercase letter that follows an all uppercase word not ending in T or I is split incorrectly.

[epage]

Closing in favor of #745. If there is a reason we should keep these separate, let us know!

[mkatychev]

Apologies for not finding the duplicate 🙇 , I'll post my test cases there.