Linear algebra tuning for nfloat + cmpabs

[fredrik-johansson]

• Implement gr_cmpabs properly for nfloat_complex, acf, acb
• Tuned threshold values for nfloat and nfloat_complex linear solving

Speedup nfloat vs arf for solving $Ax = b$ using gr_mat_nonsingular_solve (uniform random entries):

prec \ n
           2     4     8    16    32    64   128   256   512  1024 
    64  1.570 2.264 3.115 2.882 3.455 1.780 1.679 1.563 1.478 1.354 
   128  1.496 2.046 2.719 2.702 3.169 1.789 1.643 1.498 1.396 1.262 
   192  1.542 2.191 3.191 3.668 3.466 2.375 1.903 1.638 1.456 1.326 
   256  1.426 1.981 3.292 3.370 3.000 2.186 1.796 1.585 1.365 1.274 
   320  1.472 1.824 2.366 2.398 2.061 1.832 1.704 1.445 1.304 1.218 
   384  1.394 1.803 2.317 2.362 2.013 1.849 1.692 1.455 1.301 1.224 
   448  1.326 1.737 2.251 2.295 2.095 1.907 1.805 1.514 1.332 1.219 
   512  1.455 1.708 2.169 2.347 2.095 1.972 1.835 1.504 1.325 1.233 
   576  1.423 1.720 2.210 2.301 2.031 1.898 1.661 1.413 1.267 1.185 
   640  1.327 1.487 1.600 1.731 1.709 1.797 1.728 1.382 1.250 1.171 
   768  1.326 1.434 1.672 1.645 1.791 1.871 1.718 1.404 1.275 1.183 
  1024  1.398 1.476 1.588 1.691 1.846 2.028 1.789 1.480 1.321 1.206 
  1536  1.308 1.444 1.468 1.544 1.718 1.835 1.714 1.436 1.287 1.175 
  2048  1.217 1.317 1.359 1.494 1.665 1.793 1.600 1.410 1.264 1.174 
  2560  1.202 1.282 1.388 1.525 1.735 1.927 1.680 1.469 1.299 1.195 
  3072  1.169 1.262 1.314 1.477 1.654 1.831 1.594 1.426 1.281 1.179 
  4096  1.145 1.223 1.280 1.423 1.607 1.780 1.543 1.391 1.260 1.178 

Speedup nfloat_complex vs acf:

           2     4     8    16    32    64   128   256   512  1024 
    64  1.806 2.333 2.259 2.074 1.899 1.713 1.627 1.641 1.540 1.304 
   128  1.601 2.141 2.183 2.345 2.040 1.874 1.806 1.684 1.472 1.313 
   192  1.901 2.143 3.296 3.693 3.077 2.944 2.777 2.102 1.642 1.580 
   256  1.678 1.990 2.608 3.152 2.714 2.657 2.568 2.178 1.623 1.502 
   320  1.555 1.731 2.429 2.694 2.566 2.641 2.702 2.191 1.634 1.495 
   384  1.622 1.740 2.170 2.414 2.354 2.451 2.638 1.792 1.701 1.498 
   448  1.533 1.734 2.331 2.577 2.436 2.688 2.887 2.407 1.852 1.612 
   512  1.570 1.776 2.033 2.288 2.252 2.533 2.863 2.382 1.857 1.595 
   576  1.504 1.757 2.399 2.375 2.360 2.520 2.525 2.024 1.737 1.526 
   640  1.365 1.446 1.604 1.753 1.956 2.229 2.460 1.997 1.720 1.507 
   768  1.392 1.492 1.800 1.797 2.023 2.333 2.383 1.963 1.687 1.501 
  1024  2.965 1.786 1.881 2.073 2.293 2.646 2.482 2.049 1.724 1.542 
  1536  2.283 1.770 1.832 1.990 2.240 2.418 2.281 1.921 1.674 1.541 
  2048  2.216 2.100 2.064 2.135 2.326 2.534 2.239 1.937 1.636 1.512 
  2560  1.877 2.003 1.906 2.062 2.291 2.544 2.319 2.013 1.750 1.557 
  3072  1.879 1.778 1.797 1.967 2.228 2.455 2.249 1.945 1.720 1.544 
  4096  1.401 1.464 1.597 1.790 2.067 2.327 2.542 2.268 2.013 1.802