add basic test to ensure getSamples is producing unique samples

[dehann]

Came across a new case with new unrollHypo where only first smpid is used in the construction of the lambda. Easiest to see during approxConvgetSamples.

Likely requires a back port if v0.19 is effected

[dehann]

My case was local to developing a fancier getSample. Punchline is this mistake is easy to make, so perhaps a test will be good.