This is a reminder that as a followup to #9139, we want to recode the LongFloatFormat formatter in arrayprint.py so that it behaves more like the FloatFormat, for example it should now omit whitespace in the sign position if possible.

Ideally would be done before 1.14 release.

Implementation idea: Probably we should try to re-use the FloatFormat code, eg by subclassing it. The only part of FloatFormat we need to change is to use C's printf instead of python's printf-like formatter, since only the former supports the l specifier. That means about 3-4 lines need to change in FloatFormat to adapt it to LongFloatFormat. It looks like we may want to use the C-level format_longdouble function.