reactive.Value should ignore spurious sets

In R, setting a reactive value that is identical() to the existing value, results in the set being ignored. In Python, the same logic is implemented using is (object identity equality). This is different than value equality in the case of even some base types (strings, floats) and certainly any composite types.

However, using == is problematic also; for numpy and pandas, it means elementwise comparison and returns an iterable of bools. One approach would be to iterate over those bools, but that wouldn't detect differences in some types of metadata (e.g. pandas indices, maybe?).

Currently, the only place where we absolutely need equality comparison (that I know of) is the forthcoming reactive.poll. We could change reactive.value to never dedupe (or to dedupe only if opted in by providing a comparison function) and demand that reactive.poll user-provided check functions return simple data types.