#2876: Improve PyGlobus situation

PyGlobus installation fails:

* if g++ isn't installed (that causes Globus-Core to fail, failure is
non-obvious)
* If python-dev isn't installed

While this might seem low priority, it causes constant headaches because
we get frequent support email about it. How can we fix it?

A few ideas:
* We already do a check to make sure that "enough Python" is
installed. Apparently that
check isn't good enough to catch python-dev.
* Should Globus-Core fail if g++ isn't installed? But fail peacefully:
installation
could continue, with but a warning in the post-install/README.
* Could PyGlobus fail with just a warning, instead of causing the
installation to fail?
Most non-LIGO users don't care much about PyGlobus.

Crazy idea: we rewrite the build of PyGlobus to be in a script. Maybe we
invoke the script from the .pacman file, and tell people "if it fails,
try again later after you fix the following problems..." Maybe we never
invoke the script, and just make people run it if they really care.

I'm not sure what the right solution is, but I'm so sick of support
tickets about it that I really wnat to fix it before VDT 1.8.1