ValidatorDef should support phetioType

[zepumph]

Over in #257 we thought this was the best way forward. Then we don't have to do a special case for the phetioType like we do in this sha here:

axon/js/Action.js

Lines 201 to 210 in 9e7e33a

	for ( let i = 0; i < this.parameters.length; i++ ) {
	const parameter = this.parameters[ i ];
	if ( parameter.containsValidatorKeys ) {
	validate( arguments[ i ], parameter );
	}
	// valueType overrides the phetioType validator so we don't use that one if there is a valueType
	if ( parameter.phetioType && !this.parameters.valueType ) {
	validate( arguments[ i ], parameter.phetioType.validator );
	}

[zepumph]

Above I added support for phetioType as a member of ValidatorDef. This was largely straightforward. I also added tests for this and found/fixed a bug in ArrayIO's validator.

The main piece I would like reviewed here is in Property.js. Up until this point, validator keys have been passed straight to Property, but have been used to validate the value the Property is set to, not the Property itself. But when adding support for phetioType, we were passing a phetioType like PropertyIO.<BooleanIO> into a validator setup to validate boolean, not Property.<boolean>. To get around this, I am no longer passing Property's options into validator, but instead I am picking out the validator keys into a validator var, and then overwriting the phetioType to be the parameterType of the phetioType (so BooleanIO instead of PropertyIO.<BooleanIO>). The main concern is that this adds one more object (to hold the validator) per Property instance. I do not know another way around it.

@samreid please review these changes, perhaps together?

[samreid]

b55a3bf seems reasonable. To alleviate our concerns about memory, why not only allocate _.pick( options, ValidatorDef.VALIDATOR_KEYS ) if assert is truthy?

[zepumph]

How about just

     // Do not allocate the new object when assert is not enabled
      const validator = assert ? _.pick( options, ValidatorDef.VALIDATOR_KEYS ) : options;

That way we only make a new object if assert is false. What do you think?

[samreid]

That sounds good @zepumph. I also took that same idea a step further in the commit. Can you please review?

[zepumph]

Oh that is so very nice. How do you like my minor tweak to remove duplicated documentation?

[samreid]

Looks great, thanks! I'm ready to close if you are.