Add reduced Gaussian grid targets for interpolate

[sandorkertesz]

The available target grids for interpolate() are always (with some exceptions) regular lat-lon grids. This issue is addressing the question of adding reduced Gaussian grids as targets.

Proposal (targets are still to be added, exsisting ones in bold):

Source: HEALPix (nested):

Source	Target
H1024
H512
H256
H128
H64
H32
H16
H8
H4
H2

Source: octahedral reduced Gaussian

Source	Target
O2560
O1280	N320
O1024
O640
O512
O400
O320
O256
O200
O160
O128
O96
O80
O64
O48
O32

Source: reduced Gaussian (non-octahedral)

Source	Target
N1280
N1024
N640
N512
N400
N320
N256	O400, O1280
N200
N160
N128
N96
N80
N64
N48
N32

Source: regular lat-lon

Source	Target
0.25x0.25	N320

[sandorkertesz]

@iainrussell , @pmaciel what do you think?

[pmaciel]

I'm not sure we should add a big variety of grids, this could create quite a burden, now and in future releases, for no immediate value except for the exceptional user.

But certainly some are quite important. I think N320 is the format of an older reanalysis which still sees quite some usage (ERA Interim?) I might be completely off. Maybe we should ask User Services what are the popular datasets, or CDS (but those should all be regular ll, except EFAS)

[juntyr]

I recently got the error No matrix found! in_grid={'grid': 'O200'} out_grid={'grid': [1, 1]} method='linear' on v0.3.2 when trying to quickplot the https://apps.ecmwf.int/ifs-experiments/rd/guxf/ dataset

[pmaciel]

@sandorkertesz I don't know if this up-to-date. But I think the H1204 above is meant to be H1024, and possibly in nested ordering (possibly the other HEALPix with #nodes = 12*2^Nside as well)

[sandorkertesz]

I recently got the error No matrix found! in_grid={'grid': 'O200'} out_grid={'grid': [1, 1]} method='linear' on v0.3.2 when trying to quickplot the https://apps.ecmwf.int/ifs-experiments/rd/guxf/ dataset

@juntyr, if it still a problem for you can I ask you to create a separate issue for it in earthkit-regrid?