Allow Delegate parameter to change source and support a None source

[jenshnielsen]

There are a few open questions. Especially around how we handle traits such as gettable/settable/snapshot_get being inherited from the source parameter or alternatively set specifically on the Delegated parameter

[codecov]

Codecov Report

Merging #2005 into master will increase coverage by 0.04%.
The diff coverage is 84.61%.

@@            Coverage Diff             @@
##           master    #2005      +/-   ##
==========================================
+ Coverage   71.17%   71.21%   +0.04%     
==========================================
  Files         149      149              
  Lines       19541    19607      +66     
==========================================
+ Hits        13909    13964      +55     
- Misses       5632     5643      +11     

[astafan8]

as per my memory, _set_from_raw_value was made with an assumtion that it will be used in a "get" context (as opposed to "set"). this is why in _BaseParameter, _set_from_raw_value
validates the value only if _validate_on_get is True. I don't think that this assumtion is a good one. So either the _validate_on_get condition needs to be gone but then we need to be careful about validation values in group parameter case, or add the _validate_on_get condition below for consistency just "for now":

if self._parameter._validate_on_get:
    self._parameter.validate(value)

what do you think?

[jenshnielsen]

Actually this never matters. It calls set on the base parameter which will always do a validate. In anycase it should validate against the value and not the raw_value. So lets perhaps just remove this validator