cf_math: remove matrixcopy/vectorcopy in favor of memcpy()

[whoenig]

• memcpy() should be faster than an elementwise copy
• memcpy() is easier to read in the code, as the definition is well
  known

[jonasdn]

Could you rebase this on master?

Something like:

$ git rebase origin master
$ git push -f IMRCLab cf_math_cleanup3

[whoenig]

Could you rebase this on master?

Done!