Provide more builtin CustomChecks

[JelleZijlstra]

Ideas:

• NoAny: disallow Any in the passed in type. Could be either "shallow" (disallow only Any itself, and VariableNameValue) or "deep" (recurse into the Value with .walk_values() and error if there's Any anywhere).
• RestrictedLiteral: like LiteralOnly, but adds a callback function that can reject a value. For example, Spark jobs can take a string representing how much memory they accept in the form 123g. The callback could check that the passed in value is a string consisting of a number plus g. We could allow the function to return a bool, a str (error message) or an ErrorMessage object.

[JelleZijlstra]

We could also provide a JSONCompatible check that recurses into values and checks they are JSON-compatible (float, int, null, string, bool, lists or dicts of other JSON-compatible things).