The contract handlers that need the database get a reference to it in
the context. I don't need to pass another db reference to them. Fix
this.

The protocol folder heralds back to BTS, when we drew a distinction
between "protocol objects" and "implementation objects." This
distinction was never really realized, and definitely isn't being
maintained now, so do away with the vestiges. It's nothing but a
potential point of confusion now.

Throw more accurate exception types during trx validation rather than
merely assert_exception's. Also, add a message to the exception
explaining what went wrong.

Previously it was just being done in validate_tapos, which was a legacy
quirk as bts would validate expiration and tapos together. At least now,
there's no overlap between these checks, so they can be split into
separate functions.

Create a new read-write chain API with push_block and push_transaction
calls

The system contract C++ implementation is now in a plugin rather than
directly on database.

Note that testing_database installs this implementation on the testing
database, even though the test framework does not use appbase or appbase
plugins.

This is just an empty template plugin which will serve as a starting
point for new plugins.

Also add a helper script, make_new_plugin.sh, which copies the template
plugin to create your new plugin and automatically renames all of the
files/directories/classes/binaries/whatever. Invoke like so:
./make_new_plugin.sh new_plugin
Afterwards, the new_plugin folder will be ready to go.

The simulated network was recursively calling push_block when
propagating blocks, which caused a lock timeout. Fix this issue by not
pushing a new block to the database that created it.

Also, replace the currently_propagating_block flag with boost's slot
blocker.

Rename: check_tapos -> should_check_tapos
Rename: check_for_duplicate_transactions ->
should_check_for_duplicate_transactions

Since these methods return whether the named action should be taken or
not, rather than taking the named action, prefix their names with should_
so it's clear they're querying and not acting.

Add get_block call to chain read_only api

The beginnings of the chain API is now much closer to the design
described in #14

- make account names, permissions, and message types fixed strings
- move sys_contract implementation to separate cpp file
- disable deprecated warnings on OS X
- define authority class

- absent signature verification

Only one call, but that call works!

Replace the C-style typless macros in config.hpp with C++-style const
definitions.

All tests now pass again. :)

Test of the recent slots filled are now working correctly.

Yeah, P2P works now, as far as I can tell. Nodes connect, sync, and stay
in sync. Not rigorously tested yet, but the basics are there.

Dunno what happened, but I don't have time to fix it right now.

Fix the stuff that broke in merge.

Define a new p2p_plugin as the adaptor between appbase and the old P2P
library, and wire things together. This is mostly working, but for some
reason peers won't stay connected to sync, but instead just drop their
connection (with a TCP reset, no less) right before syncing, without
bothering to log about why. I suspect the failure is at a lower level. :
(