Ensure all functions correctly handle Float16

It was noted that `gamma` and `lgamma` did not handle `Float16` in #131, which was subsequently fixed in #132. I'm opening this issue to track correct handling in the remaining functions exported by this package. Please feel free to check off as appropriate.

- [x] airy
- [x] airyai
- [x] airyaiprime
- [x] airyaiprimex
- [x] airyaix
- [x] airybi
- [x] airybiprime
- [x] airybiprimex
- [x] airybix
- [x] airyprime
- [x] airyx
- [ ] besselh
- [ ] besselhx
- [ ] besseli
- [ ] besselix
- [ ] besselj
- [ ] besselj0
- [ ] besselj1
- [ ] besseljx
- [ ] besselk
- [ ] besselkx
- [ ] bessely
- [ ] bessely0
- [ ] bessely1
- [ ] besselyx
- [x] beta
- [x] cosint
- [x] dawson
- [x] digamma
- [x] erf
- [x] erfc
- [x] erfcinv
- [x] erfcx
- [x] erfi
- [x] erfinv
- [x] eta
- [x] gamma
- [ ] hankelh1
- [ ] hankelh1x
- [ ] hankelh2
- [ ] hankelh2x
- [x] invdigamma
- [x] lbeta
- [ ] lbinomial
- [ ] lfact
- [ ] lfactorial
- [x] lgamma
- [x] polygamma
- [x] sinint
- [x] trigamma
- [x] zeta