download datum://id and datum://alias files, run execution locally, using project data files

[ccschneidr]

As discussed in the Continental support channel, we need the possibility to download datum:// input files to run an execution locally. This pull request adds this feature to valohai-utils' download.py.
Using this we were able to run a complete execution, locally.

To test, perform the following steps

• log in to valohai with vh login
• link a project, if you want to use aliases with vh link
• execute python snippet (make sure to replace my_models_alias and 01941935-832f-16d4-af69-6a9bf6bf4df6 with existent data)

import valohai

valohai.prepare(
    step="predict",
    default_inputs={
        "model-http": "http://example.com",
        "model-alias": "datum://my_models_alias",
        "model-id": "datum://01941935-832f-16d4-af69-6a9bf6bf4df6",
    },
)

print(valohai.inputs("model-http").path())
print(valohai.inputs("model-id").path())
print(valohai.inputs("model-alias").path())

[hylje]

Thank you for the PR, I will adopt the branch to add improvements and add automated test cases.

[memona008]

LGTM! ❇️ some nits though