Optimize Hash#select(Enumerable) and #merge!(Hash, &)

[HertzDevil]

These two defs call has_key?(k) followed by [](k), performing a double lookup. This PR turns them into find_entry(k).

The optimization is quite noticeable when the keys are large structs, because their hash functions are more complex. I don't have a comprehensive benchmark but the speedup for #select is ~15% for UInt64 keys or ~60% for UInt8[128] keys.

[sdogruyol]

Thank you @HertzDevil 🙏

[HertzDevil]

By the way there is no public API for this when the value type is nilable, i.e. when #[]? cannot distinguish a nil value from the absence of a key. The closest thing is find { |k, _| k == key }, but it has much poorer performance and doesn't account for @compare_by_identity and @block.

[asterite]

Isn't that fetch with a block?

[HertzDevil]

#fetch yields only when the key is not present. Here the desired behavior is to do something only when the key is present.