Populate WorkflowTask.args based on defaults in WorkflowTask.task.args_schema

[tcompa]

(based on #638)

Ref:

• Default WorkflowTask arguments fractal-web#127

[tcompa]

This requires us to understand the structure of the pydantic-generated schemas (see examples in fractal-analytics-platform/fractal-web#82), so that we can reliably extract defaults, including those for nested objects.

When a Task is imported into a Workflow, as a WorkflowTask, its args attribute is filled based on the defaults in the schema.

Other than that, the args attribute can always be modified with the usual PATCH endpoint for WorkflowTask, and how it was created in the first place (from a file, in the CLI client, or from a schema, or from the combination of the two) should not matter.

Note: this issue overrides #592

[tcompa]

Here is a snippet that does part of what is needed (e.g. it does not recognize a None default):

from pydantic import BaseModel
from pydantic import Field
from devtools import debug
from typing import Optional


class TaskArguments(BaseModel):
    arg1: int = 1 
    arg2: int = Field(default=1)
    arg3: int
    arg4: Optional[int] = None
    arg5: Optional[int] = Field(default=None)


schema = TaskArguments.schema()
debug(schema)

args = {}
for arg_name, arg_dict in schema["properties"].items():
    debug(arg_name, arg_dict)
    try:
        args[arg_name] = arg_dict["default"]
    except KeyError:
        pass

debug(args)