purrr breaking changes

`map_chr()` no longer coereces from logical, integer, or double to strings.

`every()`, `some()`, and `none()` now require that `.p` return logical scalar `TRUE`, `FALSE`, or `NA`. Previously, `NA` was allowed to be a non-logical `NA`, and would be coerced to a logical `NA`.

`modify()` and variants are now wrapping `[[<-` instead of `[<-`. This change increases the genericity of these functions but might cause different behaviour in some cases. For instance, the `[[<-` for data frames is stricter than the `[<-` method and might throw errors instead of warnings. This is the case when assigning a longer vector than the number of rows. `[<-` truncates the vector with a warning, `[[<-` fails with an error (as is appropriate).

`modify()` and variants now return the same type as the input when the input is an atomic vector.

All functionals taking predicate functions (like `keep()`, `detect()`, `some()`) got stricter. Predicate functions must now return a single `TRUE` or `FALSE`. This change is meant to detect problems early with a more meaningful error message.

If `reduce()` fails with this message: ``` Error: `.x` is empty, and no `.init` supplied ``` this is because `reduce()` now returns `.init` when `.x` is empty. Fix the problem by supplying an appropriate argument to `.init`, or by providing special behaviour when `.x` has length 0.

The type predicates have been migrated to rlang. Consequently the `bare-type-predicates` documentation topic is no longer in purrr, which might cause a warning if you cross-reference it.