Task graph schema validation

[amitschang]

Attaching a schema which describes necessary and produced columns of each task could aid in pre-execution validation. For example:

• Task A takes in "id" column, produces "data" column
• Task B takes in "id" and "data", produces "measurement"
• Task C takes "measurement", produces "result"
• Task D takes "measurement", "data", produces "another_result"

These could be executed in order [A, B, C, D] or [A, B, D, C], but not [A, D].

Since the aim is to validate a possible graph and enable plugging tasks in varying arrangements, we cannot define a full schema for tasks, only the necessary input columns and the columns produced, so existing schema definitions and validation tools might not fit.