NFT Module - Use Store Indexes

[rigelrozanski]

Stemming from

• NFT Module cosmos-sdk#4209 (comment)
• NFT Module cosmos-sdk#4209 (comment)

---

Eliminate structs IDCollection, Collection and replace their functionality with index lookups

The NFT design should be restructured to eliminate the:

• Owner struct
• IDCollection struct

We should instead by applying the use of secondary indexes similarly to how the staking module stores references to the same validators by: operator-address, consensus-address, power, etc. (see example https://github.com/cosmos/cosmos-sdk/blob/master/x/staking/keeper/validator.go#L70)
(see spec: https://github.com/cosmos/cosmos-sdk/blob/master/docs/spec/staking/01_state.md#validator)

This way we should be able to easily create functions on the NFTKeeper to get the NFTs efficiently without requiring these extra object. For example I'd suggest the keeper look like:

type NFTKeeper interface {
    GetNFTByID(string) NFT
    GetCollectionByDenom(string) []*NFT
    GetCollectionsByOwner(sdk.AccAddress) []Collection
}

Or something along those lines.

---

Additionally this means that all this Collection back-updating logic can be removed (see cosmos/cosmos-sdk#4209 (comment))

[rigelrozanski]

CC @fedekunze @alexanderbez @okwme