Add resampling of time series

[kylebrandt]

In order to do binary operations (e.g. $GA + $GB) with both variables are time series, the time stamps of points between the two series will need to line up.

Roughly, this will be a new "GEL command" ("Add GEL" in the UI), so in addition to the current "Math" and "reduce" there would be "resample".

Pandas is a good system to look at for terminology and different types of resampling, this blog post seems to cover the area well: https://towardsdatascience.com/using-the-pandas-resample-function-a231144194c4

[kylebrandt]

Notes form discussion with @papagian :

The use case of resampling series to do math (e.g. $GA + $GB) means that many of parameters for the resampling operation naturally would come from other series.

• The default start time would likely be the earliest of all the series one wants to use in the operation, and the end time the latest of all the series the in operation.
• One might want an "auto-window" that is something like the smallest interval found in all the series.

Therefore, in the long term resample will probably be an operation that takes in multiple series and outputs a multivariate series. For now though it is okay to work on single sets of series until we have an idea of how multivariate series will work which its own issue: #13

[papagian]

@kylebrandt I agree that for a more fine grain resampling this approach would be the best.
And in that case maybe there won't be need for a separate Resample command and the resampling could be done internally before executing a Math expression.