Optimisation methods above -O0 break accuracy of evaluator unit test

[nedtaylor]

Compiler optimisation methods seem to bring in floating point rounding errors.

This is seen most clearly at line 2315 of mod_evolver.f90 in commit 9ceaa29. Here, the built-in sum() function provides different results to the 6th decimal place.

This rounding error them, at least seems to, be compounded in the evaluator module, where these small numbers are particularly important for accurately comparing viability of points.

For now, the fix is to not compile using optimisation flags.

[nedtaylor]

This might not be an issue. The placement sites are still valid ones, just that the order is not maintained. But that order is not fundamental.