Fix reorganization crash #2876

This is a potential fix for the bug that causes several miner nodes to crash a few days ago.

The problem: when a block is disconnected, the transactions in it are attempted to be moved to the mempool; if some of these are non-standard, this will fail. If there were transactions in the mempool already that depended one or more of those, they become orphan transactions - which breaks the mempool's consistency assumption.

In addition, this pull request adds a CTxMempool::check() method which runs an extensive consistency check on the mempool, and is called from several places when -checkmempool is passed (a non-documented flag).

Small update: made -checkmempool default in regtest mode.

Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/97bf6a7aef976cee0200c11f3915b4d7c0f33c10 for binaries and test log. This test script verifies pulls every time they are updated. It, however, dies sometimes and fails to test properly. If you are waiting on a test, please check timestamps to verify that the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/ Contact BlueMatt on freenode if something looks broken.

ACK; logic looks good, compiles and runs on OSX.

Rebased.

ACK. I think we should pull this, but I also think pulltester should gain a reproduction of it. @bluematt where are we keeping issues for the pulltester?

Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/2461aba1ac4e0673ed0568713b48e4bb37c8b041 for binaries and test log. This test script verifies pulls every time they are updated. It, however, dies sometimes and fails to test properly. If you are waiting on a test, please check timestamps to verify that the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/ Contact BlueMatt on freenode if something looks broken.