Implement Amazon S3 bucket source

[sandorkertesz]

Still work in progress.

The new source for an S3 bucket can be used like this:

import earthkit.data

# endpoint="s3.amazonaws.com"
bucket_name = "ecmwf-forecasts"
key = "20240111/00z/0p4-beta/oper/20240111000000-0h-oper-fc.grib2"

r = {"bucket": bucket_name, 
     "objects": [
         {"object":  key} 
          ],
   }

ds = earthkit.data.from_source("s3", r, stream=False, anon=True)
ds.ls()

More examples are available at: https://earthkit-data.readthedocs.io/en/feature-s3/examples/s3.html

• Multiple buckets and objects can be used
• A single part can be specified for an object as

"objects": [
         {"object": key, "start": 0, "range": 438714} 
          ],

• The default endpoint is s3.amazonaws.com. Other endpoints can be specified in the request as:

r = {"bucket": bucket_name, 
      "endpoint": "my_endpoint",
     "objects": [
     ....

• The stream option controls if the data is read as a stream or downloaded to a file. The default is stream=True
• The anon option controls whether if it is an anonymous access or AWS credentials should be used. The default is anon=True. Handling the credentials requires the aws-requests-auth and botocore packages

[corentincarton]

Tagging @chpolste as he could be interested by this for CLIMAAX.