Improvements to WindRose resampling

[misi9170]

@Bartdoekemeijer noted a few place for improvement in the WindRose class, in particular:

1. Potentially unexpected behavior of the resample_wind_rose() method, which returns a new instantiation rather than modifying the current object
2. Unexpected default behavior of plot_wind_rose(), which resamples the data provided by default so makes it ambiguous which bins were specified (also, it turns out that providing wd_step=None to avoid resampling causes plot_wind_rose() to fail).
3. Minimal support for resampling to a higher bin resolution

This PR addresses those issues.

• inplace option provided for resample_wind_rose() to modify the current object (defaults to False at the moment to retain former operation, but we could consider changing the default to True, which may be more intuitive. However, defaulting to False is more pandas-like).
• Default for plot_wind_rose() is now not to resample
• Rename plot_wind_rose() and resample_wind_rose() to plot() and resample() for consistency across WindRose and WindTIRose() (and for simplicity)
• Enhanced resampling to a higher bin resolution (@paulf81 , I'm not sure that this is going to work with the current method of creating a TimeSeries and calling to_WindRose---any ideas here?)*
• Map changes onto WindTIRose @paulf81

* We may need to be careful here, as there are several ways we could interpolate the data, so we should take a conservative/simple approach (but less conservative than the current approach, which is simply to assign extra 0 frequency bins)

[paulf81]

@misi9170 and @Bartdoekemeijer (and @ejsimley if you time?) could you check quick what you think of the commit I just pushed up to address the point above on enhanced resampling for upsampling? You'll see I made the following changes:

• I changed the name of resampling to aggregating which I think more accurately captures what it does
• I set a check that aggregating can only go to larger bins
• I added a new function resample_by_interpolation, which works like the interpolate function in v3 FLORIS.

I'll need to next follow-through on the code if you like it in principle, or can change things. But next steps will be:

• Make matching changes to WindTIRose
• Update existing tests and maybe add a couple tests
• Update examples and perhaps documentation (although that is in another pull request)

[misi9170]

@paulf81 maybe here, we should print what the current step is, which might help them set it correctly?
something like:

current_ws_step = self.wind_speeds[1] - self.wind_speeds[0]
if ws_step < current_ws_step:
    raise ValueError("ws_step provided must be at least as large as the current ws_step ({0:.2f} m/s)".format(current_ws_step)

[paulf81]

Sounds good, making the change...

[misi9170]

UserWarning("Unknown interpolation method: '{:s}'. Available methods are 'linear' and 'nearest'".format(method))

[paulf81]

fixed as well

[misi9170]

Actually what is a "UserWarning"? does it cause an error? If not, what happens next?

[paulf81]

Oof, this should be a ValueError, switching

[misi9170]

@paulf81 I like your suggestion above!

[ejsimley]

@paulf81, I reviewed the WindData aggregate and resample_by_interpolation methods and like the general idea of splitting the resampling into those two functions and like how they're implemented too.

[paulf81]

Ok @misi9170 , assuming all green checks above I think this is all good to go. The one hedge is I did update the TiRose resample -> aggregate (mostly a name change but added checking that we aren't upsampling) but didn't add an interpolate method here, fearing the 3D interpolation and being short on time, can come back later for this?