Evaluate runtime type validation libraries

[eaton]

Currently we’re using a variety of jank methods to validate and transform data structures, particularly:

• Option objects for Spider, Worker, Project, EnqueueUrls, and similar classes that operate on complex config
• Storage entities’ creation data, and JSON-compatible serialization forms
• Optional bags of data added to storage entities by project-specific code (i.e., a “content” struct added to some “resource” entities)

Our current half-baked approaches include:

• Weak typechecks like isVerticeData
• Ad-hoc custom validation (see various entity constructors)
• Ad-hoc use of the ‘is’ library
• Ad hoc use of class transformer functions during hydration/dehydration

We want something that can validate the structures of large, extensible structures; use the same definitions to check/downcast/coerce/assert, and spit out type definitions in addition to validator objects. The coercion case in particular is important; in some situations we support unserializable types and need a transform step, or support option aliases that need mapping before they’re used to instantiate things,

candidates atm include joi, ow, and superstruct.