Drop traffic from docker0 bridge to weave bridge

[bboreham]

Fixes #620, as discussed in comments there. All smoke tests still pass.

[bboreham]

Hang on, need to amend the smoke test to check that this actually works

[bboreham]

OK, have extended the smoke test - had to kludge the exec_on thing a bit to get the negation in the right place.

[rade]

Why is the kludge necessary? Would wrapping in a sh -c execution do the trick?

[rade]

Alternatively, assert_raises "exec_on $HOST1 $PING $C1" 1 should work. No need to create a helper function for it, since there is just one call site.

[rade]

Except once you've rebased from master there will be three call sites, so an exec_on1_fail would probably still be a good idea.

[rade]

Hmm, there's a problem with both your existing exec_on1_n and my proposed exec_on1_fail - they can fail (and hence report success) for entirely the wrong reasons. That's why I went for the 'negate the result of the specific check we expect to fail' approach elsewhere. So perhaps a sh -c is the way to go after all.

[bboreham]

Changed to sh -c and rebased

[bboreham]

Annularized

[rade]

I've just a) committed a change for simplicity and consistency.