Fix TokenIndexes MultiIndex

[humanalgorithm]

The Multiindex for TokenIndexes works by accident and not by design

The key on a token index should be a String and not an Addr
The conversion of Addr to String is why existing setup does not fail any unit tests

An example usage below:
If we have this data

"1": {owner: <addr1>, name: name1} 
"2": {owner: <addr1>, name: name2} 
"3": {owner: <addr2>, name: name3} 

The type structure being

Map {
String: TokenInfo {
       "owner": <Addr>
       "data": <somedata>
    }, 
String: TokenInfo {
        "owner": <Addr>
        "data": <somedata>
    }, 
    ...
 }

To uniquely identify each entry, I need to differentiate by the key, which in this case is a String
The purpose of the index, is to get token ids by providing an owner address. So,
input: <addr1>
output: ["1", "2"]

so the index is by owner and then I get back the keys associated with it ["1", "2"]

Note that an owner (addr) can have multiple token ids (string)
So index by owner, is similar to tokens "group by" owner

[shanev]

Great catch! :P

[orkunkl]

Nice one!

[orkunkl]

Good catch! I made the change back when the development speed was fast.
it is funny that,Addr is an abstraction to allow different address format validation, but in storage, it is simply an utf8 string. Lesson learned!

[the-frey]

Nice, good spot.