feat: format args and hashed expressions in math

[QuadnucYard]

Closes #22. Fixes #132 and #141.

There are some problems:

1. It is space-sensitive, and there may be some content magic on the user side. So we should not deliberately change spaces in math.
2. We cannot distinguish between sin and cases. They have the same AST, but only the latter is really a function. Therefore, we must keep commas in args and not add extra ones.

These problems mean we have to keep the structure.

Line-breaking in math is also suppressed.

[github-actions]

📊 Benchmark Performance Report

group                        base                                   pr
-----                        ----                                   --
attrs-cetz-manual            1.40     87.7±8.75µs        ? ?/sec    1.00     62.8±0.65µs        ? ?/sec
attrs-codly                  1.24   240.5±20.85µs        ? ?/sec    1.00    193.5±2.61µs        ? ?/sec
attrs-fletcher-diagram       1.28     51.0±0.39µs        ? ?/sec    1.00     40.0±6.76µs        ? ?/sec
attrs-fletcher-draw          1.29    153.5±9.98µs        ? ?/sec    1.00    118.8±2.16µs        ? ?/sec
attrs-tablex                 1.19    459.9±2.84µs        ? ?/sec    1.00    386.4±3.15µs        ? ?/sec
attrs-touying-core           1.23   293.7±10.21µs        ? ?/sec    1.00    239.7±2.63µs        ? ?/sec
attrs-touying-utils          1.25    159.0±1.68µs        ? ?/sec    1.00    127.3±2.97µs        ? ?/sec
attrs-undergraduate-math     1.31    119.7±6.89µs        ? ?/sec    1.00     91.1±1.31µs        ? ?/sec
pretty-cetz-manual           1.03    754.5±6.13µs        ? ?/sec    1.00   731.9±19.36µs        ? ?/sec
pretty-codly                 1.02  1783.9±11.88µs        ? ?/sec    1.00  1756.2±10.70µs        ? ?/sec
pretty-fletcher-diagram      1.03    511.5±7.15µs        ? ?/sec    1.00    497.6±6.80µs        ? ?/sec
pretty-fletcher-draw         1.02  1228.5±10.61µs        ? ?/sec    1.00  1205.6±12.20µs        ? ?/sec
pretty-tablex                1.02      3.4±0.04ms        ? ?/sec    1.00      3.3±0.01ms        ? ?/sec
pretty-touying-core          1.03      2.2±0.01ms        ? ?/sec    1.00      2.1±0.01ms        ? ?/sec
pretty-touying-utils         1.02  1194.9±11.22µs        ? ?/sec    1.00  1172.7±29.33µs        ? ?/sec
pretty-undergraduate-math    1.01    835.2±7.71µs        ? ?/sec    1.00    826.2±6.46µs        ? ?/sec

Generated by GitHub Actions on Fri Mar 28 02:56:10 UTC 2025

[Enter-tainer]

interestingly this pr greatly improves the performance🤔

[Enter-tainer]

snapshot LGTM

[QuadnucYard]

interestingly this pr greatly improves the performance🤔

No, just because the computation of attributes is simplified due to fewer disabling cases.