When referring to our node, we use simply `node` as opposed to
`cluster`.

Instead of propagating `rpc_processor`, use the `block_commitment_cache`
structure to get different bank confirmation levels

As ruuda pointed out, there is a deadlock if we acquire a read lock for
both `block_commitment_cache` and `bank_forks`, in ruuda words:
- We take the read lock on block_commitment_cache.
- Some other thread takes a write lock on bank_forks.
- We want to take the read lock on bank_forks, but this blocks because
  the other thread holds a write lock.
- The other thread wants to take a write lock on block_commitment_cache,
  but that blocks because we hold the read lock.
- Deadlock!

Instead, we guard the `block_commitment_cache` read lock so we don't
hold two read locks concurrently. This might lead, as ruuda also
pointed out, to an inconsistency that's known to Solana and dealt
already in the code the same way it's dealt in `rpc/src/rpc.rs`.

Should be "counter", not "count"

Also update Cargo.lock.

Down the rabbit hole from the arguments, propagating it to the
prometheus part, we add a flag so we can track multiple vote accounts.

Modify how metrics are written to account for when there's nothing to
report.
Add metrics about votes: last voted and vote balance

Change variable and function names, and add information about vote
credits

Slight change on how metrics are recorded

Co-authored-by: Ruud van Asseldonk <[email protected]>

Co-authored-by: Ruud van Asseldonk <[email protected]>

Co-authored-by: Ruud van Asseldonk <[email protected]>

With the epoch start slots and the number of slots in the epoch (and the
current slot, which we already had), we can infer/estimate:

 * Epoch progress percentage
 * Slots left until the next epoch
 * Time left until the next epoch (from slot height increase)

These are useful metrics to have about the network.

Derive a macro for `GossipStats`, a less invasive and more concise way
of writing all metrics from it.
Currently, the macro only accepts `Counter` type and will fail if used
otherwise, but we can probably expand it in the future to work with more
data types.

This function clears the gossip stats.

Since we commented the previous line, doesn't make sense to update this
variable.