FIELDEXPIRE / HEXPIRE

[NegatioN]

Hi :) Tagging @romange since I've only been speaking to you so far.

There are a few things missing in this PR, namely:

1. The generic command FieldExpire is not yet implemented.
2. There's still some mess to clean up.

I'd like some feedback on a few questions, and in general on the code:

1. Is HEXPIRE supposed to call into the more general FieldExpire after modifying the CmdArgParser, or just use the same underlying common function for manipulating field ttl?
2. Is there a way of returning an int-array in the Redis syntax to the client yet, or would I need to make that?
3. Should I remove ReplaceObj and instead use AddOrFind since we always Find the fields before modifying their ttl?

I think I also need to re-arrange the code such that I ensure the correct encoding (which supports TTL) is only enforced once for each call, and then each field is operated on afterwards. But I can't see a whole lot of code sharing between sets and hsets here, other than that they're both operating on dense_sets. I'm a c++ noob though, so feel free to correct me on this.

This PR is related to: #3027

Thanks in advance.

[romange]

Thanks for submitting this PR. A few comments:

1. This PR is quite big, so I suggest to divide it into several PRs where the first one covers only the data structures (under the core/ dir).
2. Our core classes also have unit tests. I see you are focused on HEXPIRE, which means you will need string_map to support the TTL updates. I suggest adding unit tests to string_map_test.cc for that.
3. Regarding AddOrFind - HEXPIRE never adds a field so I do not think AddOrFind is needed.

We can follow up on the questions related to the high level commands in the subsequent PRs.