Reconsider vec_ptype_abbr() for factor and complex

[krlmlr]

After switching to vec_ptype_abbr() in {pillar}, two downstream packages now fail their checks. This is mostly due to the change from "fct" to "fctr" and "cpl" to "cplx" .

type_sum() moved away from "fctr" and "cplx" just over a year ago: r-lib/pillar#80, suggested in r-lib/pillar#71. It feels inconsistent that now vec_ptype_abbr() returns "fctr" and "cplx" .

@lionel- @hadley: Can we please move to "fct" and "cpl" here to save quite a few headaches downstream?

Thanks to @nacnudus for pointing this out: nacnudus/unpivotr#24 (comment).

[hadley]

Yeah, that's fine with me. Want to do a PR?