TST: accept small error in threaded random test

freebsd and windows change x87 precision mode (fctrl bit 8 and 9) from
extended to double in child threads so the random numbers cannot be
exactly the same from master and child threads.
see gh-4909

BUG: fix array_almost_equal for array subclasses

introduced in gh-4105/ab04e1ae0e8eca717bc7e42f3b0a60c9ff764289

TST: skip clongdouble alignment checks on 32 bit arches for 1.9

turns out not only sparc is borked, skip the checks on all 32 bit arches
with too large clongdouble alignments until we have an aligned
allocator.

TST: join fft threads before getting values