Add multiple tide model support to `pixel_tides`

[robbibt]

Proposed changes

Currently, the pixel_tides function only allows us to run a single tide model at once (by default, "FES2014").

This PR updates the function to allow a user to specify a list of tide models:

tides_highres, tides_lowres = pixel_tides(
    ds=ds,
    times=times,
    model=["TPXO9-atlas-v5", "FES2014", "FES2012"],
)

Outputs are now returned as an xr.DataArray with a new "tide_model" dimension, making it easy to plot and compare different model outputs:

tides_highres.isel(time=[0, 1, 2, 3, 4]).plot(col="time", row="tide_model")

For backwards compatibility, the "tide_model" dimension is squeezed out if only one model is requested (the default).

Other minor updates:

• Added an additional test to test_coastal.py to test the existing "calculate_quantiles" functionality that is used for Intertidal Exposure.
• Switched pixel_tides to use xr_reproject from odc.geo instead of odc.algo
• Improved pixel_tides docstring and fixed up some silly mistakes

[robbibt]

To test:

1. Start up the "8XL default environment with shared drive" Sandbox environment
2. Run:

   
   import datacube
   import odc.geo.xr
   import pandas as pd
   from dea_tools.coastal import pixel_tides

dc = datacube.Datacube()

Load some data over the coast

ds = dc.load( product="ga_ls8c_ard_3", x=(122.2, 122.5), y=(-17.0, -17.3), time=("2015"), dask_chunks={}, )

Model some tides, requesting three different models

tides_highres, tides_lowres = pixel_tides( ds=ds, times=pd.date_range("2020-01-01", "2020-01-02", periods=20), model=["TPXO9-atlas-v5", "FES2014", "FES2012"], directory="/gdata1/data/tide_models_clipped/", cutoff=20, )

Plot our modelled tides from each of our models for five timesteps

tides_highres.isel(time=[0, 1, 2, 3, 4]).plot.imshow(col="time", row="tide_model")