pinning taking up lots of CPU on benchmarks

[whyrusleeping]

I was looking over the benchmarks for the dagmodifier since someone mentioned writing a large file into /ipns was slow, and i noticed that we are spending a lot of time right here: https://github.com/ipfs/go-ipfs/blob/master/pin/indirect.go#L46

We really shouldnt be marshaling keys to base58 for this...

@tv42 in your pinning stuff you were working on, you dont do any b58 marshaling, right?

[tv42]

@whyrusleeping I consider needing base58 outside of UI a design failure.

[whyrusleeping]

@tv42 good.

[whyrusleeping]

this is much better