Support lazy regridding

[stephenworsley]

Provides lazy regridding through essentially the same means as in iris (introduced in this PR SciTools/iris#3701).

[codecov]

Codecov Report

Merging #80 (f23a1d3) into main (ef0268e) will increase coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #80      +/-   ##
==========================================
+ Coverage   98.82%   98.85%   +0.03%     
==========================================
  Files          14       14              
  Lines         678      699      +21     
==========================================
+ Hits          670      691      +21     
  Misses          8        8              

Impacted Files	Coverage Δ
esmf_regrid/_esmf_sdo.py	95.83% <100.00%> (ø)
esmf_regrid/schemes.py	98.93% <100.00%> (+0.03%)	⬆️
.../schemes/test_regrid_rectilinear_to_rectilinear.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ef0268e...f23a1d3. Read the comment docs.

[jamesp]

Looks a good, a couple of minor points inline

[stephenworsley]

When this is merged, the performance test in #100 will be measuring the set up for lazy regridding but not the final realisation of the data. The realisation will also be worth measuring after this is merged in (I don't think it will make sense to add such a test before this is merged in since then it would be measuring the realisation of an already realised array).

[stephenworsley]

The following excerpt from the benchmarking CI illustrates the effect these changes have on performance:

The actual regridding step now takes up a negligible amount of time. The regridding calculations are now done during realisation of the data (previously the data was already realised, this is why the before shows up as "failed"). This calculation step now takes almost exactly half the time it did before, probably due to dask parallelisation (compare the before of time_lazy_regridding with the after of time_regridding_realisation).

[jamesp]

OK I'm happy with this. Any last changes @stephenworsley ?