Improve code coverage of unit tests

[tleonhardt]

Over time our code coverage has slowly declined from a high watermark of 97% to a current value around 87%. This has primarily been do to new features getting added which have incomplete code coverage.

We should all make an effort to increase code coverage for the 0.9.1 release. This will likely be a tricky and time consuming effort because a lot of the uncovered code is particularly difficult to unit test.

But on the positive side, our unit tests now run much, much more quickly after the migration from pyparsing to shlex. So adding more unit/integration tests should be fine from a performance perspective.

[tleonhardt]

PR #301 partially addressed this issue, in particular in relation to tab-completion of submenu commands.

[tleonhardt]

Large uncovered chunks of code include all of the following:

• argcomplete_bridge.py (@anselor)
  • CompletionFinder class (lines 143-246)
  • tokens_for_completion function (lines 45-98)
• argparse_completer.py (@anselor)
  • _format_usage method within ACHelpFormatter class (lines 670-689)
• cmd2.py (@tleonhardt)
  • ppaged function (lines 841-892)
  • history -t command (lines 3006-3031)

Covering some of this code will likely require significant mocking, but should be do-able for most of it.

[kotfu]

I've got a branch transcript_tests where I'll work on the history -t tests.

[tleonhardt]

@kotfu Thanks for working on the history -t tests.