Improve text truncation speed

[socram8888]

I was running 2.10.6 and noticed after updating to latest version that the booru was running way, way slower. I tracked it down to the truncate function, which was struggling with the thousands of words in the recent comments.

This PR severely improves the performance:

Before patch

$ time curl http://192.168.0.245/booru/index.php?q=post/list -o/dev/null
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  103k    0  103k    0     0  20205      0 --:--:--  0:00:05 --:--:-- 26273

real    0m5.258s
user    0m0.000s
sys     0m0.031s

After patch

$ time curl http://192.168.0.245/booru/index.php?q=post/list -o/dev/null
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  103k    0  103k    0     0  77359      0 --:--:--  0:00:01 --:--:-- 77317

real    0m1.413s
user    0m0.000s
sys     0m0.047s

The reason this is ultimately faster is that 1. the UTF-8 decoding is only executed once (vs multiple times as current for mb_strlen, mb_strrpos, mb_substr...), and 2. by calling mb_substr before, we will guarantee the length comparison will be bound to max ~50 characters (as opposed to strlen(a) != strlen(b) which will have to process all bytes in a and b to check their length).

[shish]

This does seem 4x faster, but FYI, if you install the mbstring library (eg php8.2-mbstring on debian), that'll be 1000x faster (I didn't realise how slow the polyfills were, maybe we should put more emphasis on having mbstring installed...)

0.69887208938599 - truncate() with polyfill
0.17476892471313 - truncate2() with polyfill
0.00040171194076 - truncate() with mbstring
0.00000390911102 - truncate2() with mbstring