Fix the 'max_tm_gxacts' value after fault injection

[whitehawk]

Fix the 'max_tm_gxacts' value after fault injection

Problem description:
After sequential execution of isolation2 tests 'standby_replay_dtx_info' and
'ao_unique_index' the coordinator's standby postmaster process together with
its children processes were terminated.

Root cause:
Test 'standby_replay_dtx_info' sets fault injection 'standby_gxacts_overflow'
on coordinator's standby, which updates the global var 'max_tm_gxacts' (the
limit of distributed transactions) to 1, but at the reset of this fault the
value of 'max_tm_gxacts' was not updated to its original value. Therefore, on
any next test that created more than 2 distributed transactions that were
replayed on the standby, the standby encountered the fatal error "the limit of 1
distributed transactions has been reached" and it was terminated.

Fix:
Set 'max_tm_gxacts' to its original value when fault injection
'standby_gxacts_overflow' is not set.

[whitehawk]

Closed in favor of GreengageDB/greengage#73