Remove all unsafe from codebase, improve parsing speed

[5225225]

Here, the toint_1/2/4 calls require both less unsafe as well as being faster, since they're specialized to the exact width they need to parse.

I ran the benchmarks and got

Before:

parse_imf_fixdate       time:   [121.04 ns 121.10 ns 121.18 ns]
parse_rfc850_date       time:   [116.55 ns 116.67 ns 116.80 ns]
parse_asctime           time:   [112.93 ns 113.04 ns 113.19 ns]
encode_date             time:   [10.981 ns 10.994 ns 11.010 ns]

After:

parse_imf_fixdate       time:   [95.967 ns 96.094 ns 96.255 ns]
                        change: [-20.734% -20.645% -20.539%] (p = 0.00 < 0.05)
                        Performance has improved.

parse_rfc850_date       time:   [87.114 ns 87.320 ns 87.559 ns]
                        change: [-25.294% -25.142% -24.972%] (p = 0.00 < 0.05)
                        Performance has improved.

parse_asctime           time:   [100.90 ns 101.05 ns 101.24 ns]
                        change: [-10.600% -10.280% -9.9340%] (p = 0.00 < 0.05)
                        Performance has improved.

encode_date             time:   [10.990 ns 11.000 ns 11.011 ns]
                        change: [+0.0702% +0.2167% +0.4071%] (p = 0.01 < 0.05)
                        Change within noise threshold.

[pyfisch]

Looks good! Can you address the comments, fix the formatting and silence clippy? Then its good to go!