std.math.big.int: Improvements to toString and setString speed

[samy-00007]

This PR is ready to be reviewed but depends on functions defined in #23848, so I am marking it as a draft until the other one is merged (hopefully).

This PR massively improve the speed of toString and setString.
For the former, a subquadratic algorithm has been added, and the quadratic one has also been improved.
For the latter, the previous implementation was, uhh, aiming for accuracy rather than speed, to put it nicely, and was therefore very easy to massively speedup.
I have also used stackFallback at 2 locations (in astgen and zongen) to reduce the number of allocations when parsing small numbers.

Graphs for toString


In the last graph (the long one), the current std implementation of toString takes 1m10 for n = 90000 on my machine.

For setString, this is the time the implementation in master takes divided by the time the new implementation takes (for a BigInt of n limbs). In short, this graph shows how many times faster is the new implementation.

Ast-check of compiler_rt/udivmodti4_test.zig comparison (master vs new):

Benchmark 1 (17 runs): /home/samy/Downloads/zig/zig-x86_64-linux-0.15.0-dev.833+5f7780c53/zig ast-check ../lib/compiler_rt/udivmodti4_test.zig
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           295ms ± 2.77ms     292ms …  302ms          1 ( 6%)        0%
  peak_rss            105MB ± 1.50MB     103MB …  108MB          0 ( 0%)        0%
  cpu_cycles          986M  ± 1.76M      984M  …  990M           0 ( 0%)        0%
  instructions       3.01G  ±  190      3.01G  … 3.01G           0 ( 0%)        0%
  cache_references   4.35M  ± 22.2K     4.30M  … 4.39M           1 ( 6%)        0%
  cache_misses       2.08M  ±  104K     1.99M  … 2.33M           2 (12%)        0%
  branch_misses      2.29M  ± 8.90K     2.27M  … 2.31M           1 ( 6%)        0%
Benchmark 2 (27 runs): stage3/bin/zig ast-check ../lib/compiler_rt/udivmodti4_test.zig
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           186ms ± 1.73ms     183ms …  192ms          2 ( 7%)        ⚡- 37.0% ±  0.5%
  peak_rss            101MB ± 2.54MB    97.2MB …  104MB          0 ( 0%)        ⚡-  4.2% ±  1.3%
  cpu_cycles          592M  ±  693K      591M  …  594M           1 ( 4%)        ⚡- 40.0% ±  0.1%
  instructions       1.77G  ±  266      1.77G  … 1.77G           0 ( 0%)        ⚡- 41.2% ±  0.0%
  cache_references   3.65M  ± 50.0K     3.56M  … 3.75M           0 ( 0%)        ⚡- 16.2% ±  0.6%
  cache_misses       1.82M  ± 92.3K     1.73M  … 2.16M           1 ( 4%)        ⚡- 12.5% ±  2.9%
  branch_misses      2.31M  ± 3.72K     2.30M  … 2.32M           1 ( 4%)          +  0.9% ±  0.2%