Use of interpolation in comparing real/synthetic simulation timeseries

> I don't think it makes sense to interpolate missing data before computing NSE, I think it makes more sense to just drop those dates and compute NSE for the rest.

_Originally posted by @cheginit in https://github.com/ImperialCollegeLondon/SWMManywhere/pull/60#discussion_r1514691199_

> My concern is that SWMM timesteps can be selected by the solver on the fly (i.e., whether it is dry or not) - and so there is a risk that there will be no perfectly overlapping datapoints if the `synthetic` and `real` simulation solvers get out of sync.
However, and probably the intuitive basis of your concern, is that if someone ramped up the `DRY_STEP` or `REPORT_STEP` values in the `.inp` file to increase speed, this interpolation could cause some significant issues.
Ultimately I think this is probably too complicated to fix here and needs its own issue, and ideally some simulations we can look at to see the balance of risks here.