split resync membership proofs

[dan-da]

addresses #183.

The comment for GlobalStateLock::resync_membership_proofs() explains:

    /// This fn performs potentially lengthy operations.  The work has been
    /// split into two: an immutable fn that prepares a batch of updates and a
    /// mutable fn that finalizes (writes) the updates as quickly as possible.
    /// In this way, we do not hold the write-lock any longer than necessary.
    ///
    /// Splitting the logic this way requires buffering updates in RAM.  We batch
    /// the updates to avoid running out of mem.  The batch size is initially
    /// set to 1000.  This is just a wild-ass-guess.  It seems high enough that
    /// most wallets will only need a single batch, but low enough that we
    /// shouldn't run out of mem if processing a huge wallet.