Floating-point rounding issues in test parameters

[rgerkin]

Test parameters (and probably observation values) have awkward representation issues (see e.g. Amplitude below):

That is from the test "RheobaseTestToRemove" on the staging server. Desired behavior would be to represent 1e10 as 1e10, not as 1.0000000000002e10. This could mean using higher precision floats or explicit representations of decimals.