Highter success proba for small blocksizes

[lducas]

In small blocksizes, I suspect the cost of preprocessing and just setting up enumeration outweight the enumeration itself. We may certainly increase the success probability in those dimension (say up to < 40, linearly derea<sing success proba from .99 down to .51).

I suggest to only do so after having switched from mpz to long ints (the above guesstimate may change after this switch is done)

[lducas]

@malb, please poke me when you've re-ran strategizer (note: to adjust my guesstimate, I can do that from strategies up to say 55, no need to waste cycle here)