set hasSpare to false at initialization of random number generator

[oscarxblanco]

This PR is a possible solution to issue #978 where the Buffer implementation for the WHITENOISE mode of the VariableThinMultipole (see #839 ) shows a cycle of 2 after repeated calls to track functions.

The bool variable hasSpare in the pcg_state_setseq_64 struct was unset at initialization, leaving the next state in an alternate True/False value from subsequent calls.

This fix proposes to set hasSpare to False at initialization. I have done a test in pyat and the sequence behaves as expected, i.e. it repeats when a seed is given.

>>> zout,*_ = thering.track(zin[:,0], nturns=5,seed=20)
>>> elewhitenoise.BufferA
array([-0.50023502, -0.71931482, -0.59066089,  0.29808669, -0.5266456 ])
>>> zout,*_ = thering.track(zin[:,0], nturns=5,seed=20)
>>> elewhitenoise.BufferA
array([-0.50023502, -0.71931482, -0.59066089,  0.29808669, -0.5266456 ])
>>> zout,*_ = thering.track(zin[:,0], nturns=5,seed=20)
>>> elewhitenoise.BufferA
array([-0.50023502, -0.71931482, -0.59066089,  0.29808669, -0.5266456 ])
>>> zout,*_ = thering.track(zin[:,0], nturns=5,seed=20)
>>> elewhitenoise.BufferA
array([-0.50023502, -0.71931482, -0.59066089,  0.29808669, -0.5266456 ])

[oscarxblanco]

Dear all, who could review this PR ?

[swhite2401]

@oscarxblanco , I believe @lfarv should review it.

[lfarv]

Well done, thanks. I hope that this will solve the problem of random generators with OpneMP. I'll check that…