This merges work done by yorickvP and timokau in #63208 and #63616 respectively.
Now the derivation builds both libtensorflow and the Python package and puts them into
different outputs.
Quite a bit of improvements were done on the top, including:
* Use official tag revision as source, not a branch;
* Use all system libraries possible (before only one was actually used);
* Move various environment variables to the derivation itself from hooks;
* Use source Python build instead of wheel build to ensure fixup hooks do their important jobs on libraries;
* And more that I forgot!
The `buildPython*` function computes name from `pname` and `version`.
This change removes `name` attribute from all expressions in
`pkgs/development/python-modules`.
While at it, some other minor changes were made as well, such as
replacing `fetchurl` calls with `fetchPypi`.
Google publishes prebuilt tensorflow whl for python 3.4, 3.5, 3.6,
but nix expression for tensorflow only supported 3.5.
This change adds support for python-3.6.
When using cuda, the rpath was set to include GCC lib version 4.9.
I am not sure what this was attempting to do, but an effect was to
prevent certain python libraries to find the correct (newer) version
of the std lib.
Also avoid mentions of any specifc version in the
propagatedBuildInputs
