feat: add Binary constraint to prevent energy losses (at negative prices)

[Flix6x]

To generate each row in my performance test, I changed line 181 in the test:

end = tz.localize(datetime(2015, 1, 4, 2))  # 8 steps
end = tz.localize(datetime(2015, 1, 4, 3))  # 12 steps
# etc..

And for the column without binary constraint I just commented out the binary constraints (lines 334-339) in linear_optimization.py.

[Flix6x]

I feel that make test should install HiGHs now. Perhaps @nhoening can consult us on how to best do that. For instance, pip install highspy could be a new install instruction under install-for-test in our Makefile. Or should we be keeping track of the version of the HiGHs pip package within our dependency files?

[victorgarcia98]

I feel that make test should install HiGHs now. Perhaps @nhoening can consult us on how to best do that. For instance, pip install highspy could be a new install instruction under install-for-test in our Makefile. Or should we be keeping track of the version of the HiGHs pip package within our dependency files?

That makes sense. What about make install-for-dev? Maybe it's a good idea to shorten the setup time.

[nhoening]

I am happy to include in install-for-test and even dev (as it's the default and installing is easy).

Tracking the version is a good thought, but I can't decide if we should do that here. Others solvers we support are not tracked by us this way. For now, I'd leave it at that.

We probably are not using intricate features of the solvers, right?

[victorgarcia98]

@Flix6x The changes that you introduced look good to me!