Add lazy loading option for the FDB source

[sandorkertesz]

This PR implements lazy loading for GRIB data read from FDB. The idea is that when lazy=True is used:

request = {
    "class": "od",
    "expver": "0001",
    "stream": "oper",
    "date": [20240603, 20240604],
    "time": [0, 1200],
    "domain": "g",
    "type": "fc",
    "levtype": "pl",
    "levelist": [500, 700],
    "step": [0, 6],
    "param": [130, 157],
}

ds = from_source("fdb", , config=config, stream=False, lazy=True)

from_source does not execute the retrieval but returns a Virtual FieldList. This involves the following steps:

• The the request for each available field is determined by using pyfdb.list()
• Create one Virtual Field for each request. These fields form a Virtual Fieldlist
• The first field is retrieved as a reference, while the other fields only contain their requests

Using any metadata queries on the Virtual Fieldlist/Fields does not involve any data retrieval, but try to use the requests for it. For metadata keys not present in the request the metadata from the reference field is used.

# No retrieval is needed for these calls
ds1 = ds.sel(param="t", levelist=500)
ds[1].metadata("param")
xr_ds = ds.to_xarray()

Any calls requiring the data values in a field trigger the actual retrieval and the resulting GRIB field will be stored in memory for each virtual field.

# This will trigger the data retrieval if data is not yet retrieved
ds[1].to_numpy()

Notes

• retrievals are performed per field. However, grouping may be more efficient
• each Virtual Field retrieves its corresponding GRIB field into memory temporarily when the values are accessed. This GRIB field is not cached and gets deleted when going out of scope (as soon as the values are returned from the Virtual Field). This behaviour should be controlled allowing options to cache it in memory or on disk.
• we could determine the individual field requests by splitting the original request instead of using pyfdb.list()

[codecov-commenter]

Codecov Report

❌ Patch coverage is 32.50000% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.82%. Comparing base (4c937d9) to head (9c5a256).
⚠️ Report is 352 commits behind head on develop.

Files with missing lines	Patch %	Lines
tests/lazy/test_lazy_fdb.py	32.50%	27 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #677      +/-   ##
===========================================
- Coverage    91.01%   90.82%   -0.19%     
===========================================
  Files          162      163       +1     
  Lines        12391    12431      +40     
  Branches       609      609              
===========================================
+ Hits         11278    11291      +13     
- Misses         932      959      +27     
  Partials       181      181              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.