WIP: myTAP implementation of pgTAP `runtests`

[sq6jnx]

... as mentioned in #48. This branch is on top of #49.

[sq6jnx]

Travis / my_prove said:

Test Summary Report
-------------------
tests/test-mytap-runtests.my  (Wstat: 0 Tests: 14 Failed: 2)
  Failed tests:  1, 1
  Parse errors: More than one plan found in TAP output
                Tests out of sequence.  Found (1) but expected (13)
                More than one plan found in TAP output
                Tests out of sequence.  Found (1) but expected (14)
                Bad plan.  You planned 12 tests but ran 14.
Files=22, Tests=1410,  6 wallclock secs ( 0.44 usr  0.08 sys +  0.10 cusr  0.10 csys =  0.72 CPU)
Result: FAIL
The command "my_prove -v -u root -h 127.0.0.1 tests/*.my" exited with 1.

The only possibility I see is to run all assertions in one plan.

[sq6jnx]

Interesting case in build 31 for MySQL 5.7 (https://travis-ci.org/hepabolu/mytap/builds/603775539?utm_medium=notification&utm_source=github_status) -- it look there is a bug which arises when using constructions like SELECT tap.ok((<some SELECT with JOINS>), ...).

It's safer to use session variables.