Add Cython CFLAGS for package optimizations (#1904)
* Add Cython CLFAGS for package optimizations
* These flags should make the code faster and
smaller
* This will affect cpython generated shared
libraries
* Use ENV to set/unset CFLAGS
* Wrap CFLAGS around pip installs
* Combine append lines
* Fix gocritic error of multiple append lines.
* Remove —ffast-math for CFLAGS
* This has a problem of stomping on float
subnormals
* This can adversely affect ML workflows
