Improve the duplication method of the stack to reuse big ints by passing
in an existing integer pool.

Improved push instructions by removing unnecessary big int allocations
and by making it int instead of big.Int

Add the object to the list of destructed objects during a selfdestruct /
suicide operation and also remove it from the list once the journal
reverts.

This commit is a preparation for the upcoming metropolis hardfork. It
prepares the state, core and vm packages such that integration with
metropolis becomes less of a hassle.

* Difficulty calculation requires header instead of individual
  parameters
* statedb.StartRecord renamed to statedb.Prepare and added Finalise
  method required by metropolis, which removes unwanted accounts from
  the state (i.e. selfdestruct)
* State keeps record of destructed objects (in addition to dirty
  objects)
* core/vm pre-compiles may now return errors
* core/vm pre-compiles gas check now take the full byte slice as argument
  instead of just the size
* core/vm now keeps several hard-fork instruction tables instead of a
  single instruction table and removes the need for hard-fork checks in
  the instructions
* core/vm contains a empty restruction function which is added in
  preparation of metropolis write-only mode operations
* Adds the bn256 curve
* Adds and sets the metropolis chain config block parameters (2^64-1)

execQueue used an atomic counter to track whether the queue had been
closed, but the checking the counter didn't happen because the queue was
blocked on its channel.

Fix it by using a condition variable instead of sync/atomic. I tried an
implementation based on channels first, but it was hard to make it
reliable.

quit now waits for the queue loop to exit.

Previously, resolve errors were being swallowed and the returned error
was a generic "not a content hash" which isn't helpful.

This updates the Resolve function to fail fast rather than only
returning an error at the end, and also adds test coverage.

CentOS has been added as a multi-machine option to the Vagrant script.
Ubuntu is still the default option. For starting the CentOS machine, use:

   vagrant up centos

mobile: add toString & rlp/json encoding for protocol types

cmd, node: support different bootnodes, fix default light port

core: fix processing regression during receipt import

common/bitutil: added data compression algorithm

travis: adapt build script to new travis VM settings

common/bitutil, consensus/ethash: reusable bitutil package

cmd, core, params: add --rinkeby flag for fast connectivity

* core: Fix for consensus test gasLimit > 2^63-1 https://github.com/ethereum/tests/blob/develop/BlockchainTests/bcInvalidHeaderTest.json#L238

* core: fix testcase for uncle gasUsage > gasLimit : https://github.com/ethereum/tests/blob/develop/BlockchainTests/EIP150/bcUncleHeaderValiditiy.json#L986

* math/big: rename TTM63m1 -> MaxBig63, + go fmt

* core: documentation

Migrate remaining flags/command to new style

cmd/faucet, cmd/puppeth: support multi-tiered faucet