Add a test that reproduce something-for-nothing fill bug

[takaaki7]

Added a test that reproduces something-for-nothing fill bug #184 after hardfork555.

To put it briefly, following process reproduces the bug.

create_sell_order(core_seller, core.amount(1), test.amount(2));
create_sell_order(core_seller, core.amount(1), test.amount(2));
create_sell_order(core_buyer, test.amount(3), core.amount(1));

It's because maybe_cull_small_order() logic is deferred for a taker order. Even when amount_for_receive() become 0 for the taker, the order might be matched with other orders in case amount_for_sale() isn't 0.
(https://github.com/bitshares/bitshares-core/blob/master/libraries/chain/db_market.cpp#L182-L203 )

As discussed in related issues, this should be fixed carefully.

[pmconrad]

Excellent! Tested to reproduce the broken behaviour.

One remark though: the test should fail, because the observed behaviour is a bug.

[takaaki7]

I thought this behavior is expected for operations in the term between hardfork555 and a next hardfork that will fix this problem. But it's too complicated and I modified this to fail.

[pmconrad]

Ok, thanks!

[oxarbitrage]

nice work, tested. merging.