========================================================================
PractRand defines several batteries of tests for RNGs.  These are 
defined as PractRand::Tests::ListOfTests and as functions in 
PractRand::Tests::Batteries that return ListOfTests objects.  When you 
are done with a ListOfTests returned by one of those functions you 
should call PractRand::Tests::Batteries::destruct_tests on it to 
deallocate the resources used by the tests.  

The predefined batteries of test in PractRand can be thought of as 
organized in a 2x3 grid.  One axis corresponds to whether the testing 
algorithms used are the normal set of algorithms or the expanded set 
of algorithms.  The other axis corresponds to the degree of "folding" 
by which PractRand concentrates extra effort on the lowest bits of 
the PRNGs output.  

This chart shows the functions used to instantiate each battery of 
tests from the grid.  Each of these functions can be found in the 
PractRand::Tests::Batteries namespace.  
                  Normal Test Set        Expanded Test Set
No Foldings       get_core_tests()       get_expanded_core_tests()
Standard Foldings get_standard_tests()   get_expanded_standard_tests()
Extra Foldings    get_folded_tests()     get_expanded_folded_tests()

The test sets are:
	Normal Test Set: (recommended)
	These tests are optimized for maximum effectiveness per cycle.  

	Expanded Test Set:
	These tests reduce per-cycle effectiveness in order to improve 
	per-bit effectiveness.  

The folding options are:
	No Folding:
	Just use the tests.  For cases where you're confident that no 
	low-bit issues could occur.  Perhaps the algorithm is provably 
	symetric in terms of bit positions, though even then folding 
	sometimes helps.  

	Standard Foldings: (recommended)
	Concentrate extra testing on the lowest bit positions.  Usually a 
	good idea.  This is done by having multiple instances of the test 
	set - the core set just tests the data, but duplicate sets work on 
	additional datastream created by skipping all bits of input except 
	for the lowest bit(s) of each integer.  These extra datastreams 
	slow down testing by about 30% and test evaluation by about 150%.  
	- Important Note: This is actually several closely related sets 
	of foldings, one for each PRNG output format (one for 8 bit 
	integers, one for 16 bit integers, 32 bit, 64 bit, and one 
	for unidentified PRNG output formats).  Which one gets used 
	is determined by examining the PRNGs metadata.  If for some 
	reason the metadata is erroneous or missing this may not be a 
	good idea.  

	Extra Foldings:
	Similar to Standard Foldings, except that instead of using 
	the metadata to identify the output format of the PRNG (and 
	thus which bits of metadata might benefit from extra testing) 
	this variant uses a much larger number of extra datastreams 
	to try to find good bits to target for every common output 
	format.  This may cause excessive amounts of memory to be 
	used on longer data streams (excessive meaning too much for 
	a 32 bit program; try using a 64-bit build instead).