Feature: `HashTable::try_insert_unique_within_capacity`

I'm experimenting with no_panic in FFI calls which use HashTable. The goal is to "prove" that the extern "C" function cannot panic, which is a nice goal because it's not allowed to.

I can use HashTable::try_reserve to avoid the panic in practice. However, that's not good enough for no_panic; if the code for panics exists when linking, then no_panic will cause the crate to fail to build. The problem specifically with insert_unique is that it calls RawTable::insert which calls self.reserve(1, hasher), which will panic out of necessity if allocation fails.

So, I propose add HashTable::try_insert_unique_within_capacity which looks like:

    pub fn try_insert_unique_within_capacity(
        &mut self,
        hash: u64,
        value: T,
    ) -> Result<OccupiedEntry<'_, T, A>, T>

And then we remove the feature gate of #[cfg(feature = "rustc-internal-api")] from RawTable::insert_no_grow, or create an equivalent method that's only pub(crate) or something. Edit: reworked to be safe and return Result instead.

This doesn't seem difficult at all--is there maybe a reason it doesn't already exist?

Start by reading HashTable::insert_unique, RawTable::insert, RawTable::insert_no_grow, and HashTable::try_reserve to understand the existing capacity and allocation paths. Add the proposed safe Result-returning API, then verify that insertion within capacity avoids the panic-producing path and reports failure through the result.