refactor: Fix new and old codechecker script issues

[nettle]

Why:
Change #199 caused a few bugs in downstream projects.
Migration to py_binary should be fixed and refactored.
Besides codechecker.bzl and codechecker_script.py also
contained old bugs.

What:

• Execute the script py_binary in the test rule, drop the symlink
• Declare the script tools in the exec configuration
• Pass one --env= per entry
• Build the environment explicitly, pass it to every command
• Simplify the script, name the arguments after the flags
• Add tests for the analyzer paths, parse exports and config env

[nettle]

Hi @furtib, @Szelethus,
I have added a few comments to highlight the most important bugfixes

[furtib]

Looks good!
I only have this tiny nit regarding target/exec.

[nettle]

Hi @furtib,
I have tested "exec" in downstream projects, and I confirm that "exec" is not just theoretically correct suggestion but also proven in practice. If we put "target" then it means we request the script to be build for execution on target platform which is wrong - all executions are done on Execution platform (see https://bazel.build/concepts/platforms).
With "target" it also does not fail since it is Python script, but "exec" seems correct. Note that we may also consider "host" (platform on which Bazel itself runs) but I think we are closer to "exec" (platform which runs compile actions).

[furtib]

In that case, if we expect the test actions to also be executed on the remote machines, then exec indeed seems to be the correct choice here.

LGTM!