Reimplement the generic fmod

[quaternic]

Redo of rust-lang/libm#536 since the repository was moved.

Full reimplementation for fmod, that should be somewhat cleaner. This was showing a decent perf gain locally, but the major improvement will come later by implementing the reduction helper with something smarter, which does involve some tradeoffs and alternatives to consider.

While applying the feedback from the original PR, I used the opportunity to go over the code again.

[tgross35]

Sorry about CI, looks like I have a few things to fix there still

[tgross35]

One question but looks great to me.

[tgross35]

Pretty great improvements!

icount::icount_bench_fmod_group::icount_bench_fmod logspace:setup_fmod()
  Baselines:                      softfloat|softfloat
  Instructions:                     1019076|1102967              (-7.60594%) [-1.08232x]
  L1 Hits:                          1021275|1105168              (-7.59097%) [-1.08215x]
  L2 Hits:                                2|1                    (+100.000%) [+2.00000x]
  RAM Hits:                              11|10                   (+10.0000%) [+1.10000x]
  Total read+write:                 1021288|1105179              (-7.59072%) [-1.08214x]
  Estimated Cycles:                 1021670|1105523              (-7.58492%) [-1.08207x]
icount::icount_bench_fmodf128_group::icount_bench_fmodf128 logspace:setup_fmodf128()
  Baselines:                      softfloat|softfloat
  Instructions:                    30701786|31329198             (-2.00264%) [-1.02044x]
  L1 Hits:                         30717679|31363321             (-2.05859%) [-1.02102x]
  L2 Hits:                                4|4                    (No change)
  RAM Hits:                              38|38                   (No change)
  Total read+write:                30717721|31363363             (-2.05859%) [-1.02102x]
  Estimated Cycles:                30719029|31364671             (-2.05850%) [-1.02102x]
icount::icount_bench_fmodf16_group::icount_bench_fmodf16 logspace:setup_fmodf16()
  Baselines:                      softfloat|softfloat
  Instructions:                       50124|84006                (-40.3328%) [-1.67596x]
  L1 Hits:                            56766|100646               (-43.5984%) [-1.77300x]
  L2 Hits:                                4|1                    (+300.000%) [+4.00000x]
  RAM Hits:                              15|17                   (-11.7647%) [-1.13333x]
  Total read+write:                   56785|100664               (-43.5896%) [-1.77272x]
  Estimated Cycles:                   57311|101246               (-43.3943%) [-1.76661x]
icount::icount_bench_fmodf_group::icount_bench_fmodf logspace:setup_fmodf()
  Baselines:                      softfloat|softfloat
  Instructions:                      151547|186691               (-18.8247%) [-1.23190x]
  L1 Hits:                           153748|188893               (-18.6058%) [-1.22859x]
  L2 Hits:                                1|1                    (No change)
  RAM Hits:                              10|9                    (+11.1111%) [+1.11111x]
  Total read+write:                  153759|188903               (-18.6043%) [-1.22857x]
  Estimated Cycles:                  154103|189213               (-18.5558%) [-1.22783x]