Enhance test coverage for already covered commands

[ejpcmac]

Rationale

Currently, test coverage is not close to 100% even in already covered commands. While 100% coverage is not an absolute goal, it would be good to understand what are the misses to take informed decisions on them.

Concept

Let’s:

• Analyse coverage misses
• Decide what to do about them

Analysis

Sources of coverage misses:

• Untested commands (some internal APIs are then not covered as a result)
• Logs
• Untested valid behaviour
• Terminal size
• Corner cases in helpers
• Corner cases in external interactions (git in some cases, file permission erorrs, …)
• Some very unlikely error cases are not tested (i.e. git commit not being callable, though git being callable has already been tested before in the execution)
• False positives (this code is tested by this test) => actually these kind of false positives are due to not executing the command to the end with rexpect.

Actions

Let’s:

• Enable tracing logs in tests
• Always execute commands to the end so that the coverage report is taken into account
• Add missing tests for observable valid behaviour
• Check whether we can manipulate the terminal size to test size-dependent code (spend max. 15 minutes looking for that and discard otherwise) => Seems like it is not possible: Allow specifying terminal size rust-cli/rexpect#119
• Write trivial unit tests for corner cases in helpers

Let’s not:

• Write tests for untested command (they will be tested in Add tests for git z init #36 and Add tests for git z update #37)
• Write tests for corner cases which do not even have a proper user error
• Write tests for very unlikely error cases (are they even testable?)