To avoid hanging test case, always mine until txpool empty

[uwiger]

A hanging test case in aecore_forking_SUITE manifested as follows (from log on dev1):

2023-03-20 10:28:41.901 [debug] <0.490.0>@aec_tx_pool:handle_info:516 [0] Msg = {gproc_ps_event,top_changed,#{info => #{block_hash => <<114,42,102,250,120,216,158,150,151,209,220,100,185,192,153,20,249,237,156,11,213,4,213,39,41,244,113,193,54,248,50,76>>,block_type => micro,height => 23,prev_hash => <<4,71,42,121,145,171,133,228,235,139,54,1,23,155,118,132,157,2,51,178,4,14,179,109,33,38,40,55,151,2,123,242>>},sender => <0.498.0>,time => {1679,308121,901649}}}
2023-03-20 10:28:41.901 [debug] <0.497.0>@aec_block_generator:handle_info:119 ignored event top_changed
2023-03-20 10:28:41.906 [debug] <0.497.0>@aec_block_generator:handle_info:119 ignored event tx_received
2023-03-20 10:28:41.907 [debug] <0.498.0>@aec_conductor:handle_add_block:1368 insert_block -> {ok,true,{key_header,23,<<27,31,57,141,53,191,129,20,78,253,20,14,29,8,98,227,94,3,3,100,226,124,151,249,137,35,2,199,185,86,54,141>>,<<27,31,57,141,53,191,129,20,78,253,20,14,29,8,98,227,94,3,3,100,226,124,151,249,137,35,2,199,185,86,54,141>>,<<43,239,74,146,79,98,161,26,199,80,4,68,99,233,43,197,194,140,95,109,131,170,69,209,17,41,63,108,253,210,54,140>>,553713663,3750491570719023204,1679308120572,6,[190,427,486,818,1217,1320,2223,2486,3118,3270,3556,4783,5875,6868,8015,8740,8837,8937,10191,10380,10802,10958,11394,13110,13412,13700,14289,15796,18576,19883,20306,20389,21041,22980,25267,25441,26841,27648,30003,30201,31137,31584],<<172,242,11,25,225,183,174,210,42,187,82,216,134,39,154,216,75,144,177,200,25,173,70,9,101,4,58,20,21,84,115,110>>,<<40,25,191,50,209,111,19,239,98,126,125,211,15,133,93,12,13,125,167,137,94,138,27,55,23,50,106,33,28,222,180,102>>,<<0,0,2,169>>,#{consensus => aec_consensus_bitcoin_ng}},[]}
2023-03-20 10:28:41.907 [debug] <0.490.0>@aec_tx_pool:enter_tx_gc:809 Adding "<<140,204,61,43,...,95,218,251,162>>" for GC at 1023
2023-03-20 10:28:41.908 [debug] <0.490.0>@aec_tx_pool:handle_call:403 [840] Msg = {top_change,#{new_hash => <<251,93,75,66,148,247,129,117,161,132,143,189,3,125,140,124,128,62,73,182,188,239,117,44,16,181,211,140,94,180,88,14>>,new_height => 24,old_hash => <<114,42,102,250,120,216,158,150,151,209,220,100,185,192,153,20,249,237,156,11,213,4,213,39,41,244,113,193,54,248,50,76>>,old_height => 23,prev_new_hash => <<4,71,42,121,145,171,133,228,235,139,54,1,23,155,118,132,157,2,51,178,4,14,179,109,33,38,40,55,151,2,123,242>>,type => key}}
2023-03-20 10:28:41.908 [debug] <0.490.0>@aec_tx_pool:handle_info:516 [5] Msg = {gproc_ps_event,top_changed,#{info => #{block_hash => <<251,93,75,66,148,247,129,117,161,132,143,189,3,125,140,124,128,62,73,182,188,239,117,44,16,181,211,140,94,180,88,14>>,block_type => key,height => 24,prev_hash => <<4,71,42,121,145,171,133,228,235,139,54,1,23,155,118,132,157,2,51,178,4,14,179,109,33,38,40,55,151,2,123,242>>},sender => <0.498.0>,time => {1679,308121,908286}}}
2023-03-20 10:28:41.908 [debug] <0.497.0>@aec_block_generator:handle_info:119 ignored event top_changed
2023-03-20 10:28:41.908 [debug] <0.834.0>@aec_tx_pool:do_update_sync_top:722 do_update_sync_top(24,23,<0.490.0>), LocalTop = 24
2023-03-20 10:28:41.908 [debug] <0.490.0>@aec_tx_pool:handle_info:516 [2] Msg = {<0.834.0>,new_gc_height,24}
2023-03-20 10:28:41.909 [debug] <0.498.0>@aec_tx_pool:garbage_collect:247 garbage_collect()
2023-03-20 10:28:41.909 [debug] <0.490.0>@aec_tx_pool:handle_cast:500 [4] Msg = garbage_collect
2023-03-20 10:28:41.909 [debug] <0.837.0>@aec_tx_pool:do_update_sync_top:722 do_update_sync_top(24,24,<0.490.0>), LocalTop = 24
2023-03-20 10:28:41.909 [debug] <0.497.0>@aec_block_generator:handle_call:83 stop_generation
2023-03-20 10:28:41.909 [debug] <0.490.0>@aec_tx_pool:handle_info:516 [1] Msg = {<0.837.0>,new_gc_height,24}
2023-03-20 10:28:42.048 [debug] <0.490.0>@aec_tx_pool:handle_call:403 [12] Msg = {peek,infinity,all}
...

The assumption is that the preceding test case, mining exactly one microblock on dev2, didn't empty the mempool, after which, the following test case would wait endlessly (= until time trap) for its mempool to become empty.

This change ensures that dev2 keeps producing microblocks as long as there is something in its mempool.