[WIP] Avoid recompiling executable if already compiled in previous tests

[RemiLehe]

By default, when performing several tests with regression_testing.py, the executable is recompiled from scratch for every test. The compilation can therefore dominate the time taken by the tests.

This PR proposes the avoid re-compiling the executable, if it was already compiled with the exact same options in a previous test. We are currently using this feature in order to same time in the Travis CI tests for WarpX.

In practice, this is done by:

• Creating a new Build folder, where compiled executable are stored.
• When running a new tests, comparing compiling options to those used in previous tests. If they match, the previous executable is retrieved from the Build folder. Note: this assumes that each executable name is unique ; i.e. that
  two sets of compiling options cannot produce the same executable name.

In order to obtain the string that contains compiling options, parts of the function build_c was extracted as a separate function.

[maximumcats]

Remi, many of the codes require builds that are specific to a given regression test even when the same compiler options are used, since they have problem-specific setup routines. Does this change break that case? If not, great. If so, can it be made an option?

[zingale]

can this be opt-in? Based on some new flag like executable_names_are_unique? Because it is not the case with our codes that changing options gives a different executable name -- for instance in Castro, the nuclear reaction network, EOS, conductivity, integrator, ... all are not reflected in the executable name.

[RemiLehe]

OK, thanks for the feedback!
Yes, I'll update the PR to make this behavior optional (deactivated by default).

[zingale]

oh sorry... I did the change from master to main, but that closed this. Can you reopen, targeting main?