Future plans for the `longdouble` argument

[spencerkclark]

In #284 @mcuntz added a longdouble argument to enable encoding times (in memory) with greater precision. The default was set to False, since otherwise it would interfere with serialization logic in downstream libraries (e.g. xarray). I just wanted to check if it is still planned to eventually change the default to True. In xarray this would be fairly straightforward to address by explicitly setting the longdouble argument to False in the time encoding logic, but before doing that I just wanted to confirm cftime's future plans.

[jswhit]

I think it would be good to change the default to True in a future release, as long as xarray supports it.

[spencerkclark]

Sounds good. I'll make that PR to xarray when I get a chance.