Add mutable hash map

[yj-qin]

This PR adds a mutable hash map implements with Robin Hood Hashing. It based on an open-addressing hash table that uses linear probing to resolve hash collisions. When removing key-value pairs, it uses the shift-back method for reordering.

[waterlens]

I feel the implementation of the hash table should be considered to reference modern implementations in Rust and Cpp. It's preferable to use integers rather than float points arithmetics to estimate the load factor. And it's not common to use a large load factor for Robin Hood hash table, considering that it will quickly be worse than a sorted array by hash values. Also for #90.

[yj-qin]

I feel the implementation of the hash table should be considered to reference modern implementations in Rust and Cpp. It's preferable to use integers rather than float points arithmetics to estimate the load factor. And it's not common to use a large load factor for Robin Hood hash table, considering that it will quickly be worse than a sorted array by hash values. Also for #90.

Integer arithmetics will be better. Maybe we can do a benchmark to find a better load factor?

[waterlens]

I feel the implementation of the hash table should be considered to reference modern implementations in Rust and Cpp. It's preferable to use integers rather than float points arithmetics to estimate the load factor. And it's not common to use a large load factor for Robin Hood hash table, considering that it will quickly be worse than a sorted array by hash values. Also for #90.

Integer arithmetics will be better. Maybe we can do a benchmark to find a better load factor?

I think it should be fine between 0.5 and 0.75. But doing a benchmark should be a better approach for this.

[peter-jerry-ye]

@yj-qin What else do you expect to change apart from the load factor?
@bobzhang Do you think we need a benchmark to find the load factor or we can merge for now to provide the functionality first?

[yj-qin]

@yj-qin What else do you expect to change apart from the load factor? @bobzhang Do you think we need a benchmark to find the load factor or we can merge for now to provide the functionality first?

I think It's ready to merge if current load factor (13/16 about 81.25%) is acceptable. We can optimize the load factor or the implementation of the hash table later.

[peter-jerry-ye]

@yj-qin If you are ready, we are merging this