Test suite failures in IDFILE tests if backslash in TMP path

Due to the way pseudo-envvar substitution is done on the test command configuration, inputs containing a backslash may be mangled. Given Windows uses backslash as a directory separator, this is a particular problem there:

    re.sub("\$BLAH", "C:\\temp", "$BLAH") -> "C:\x09emp"

If the path contains a backslash followed by a number instead of just 't' then `sre_constants.error: invalid group reference` will be raised.

The problem is fixed in r113.1.4 but this part of the lp:~gz/testrepository/misc_workarounds branch was not merged. Apparently fixing this interferes with the idea of embedding shell scripts into the configuration files as requested for bug 595295 but I can't for the life of me see how (or why that should be encouraged).

On my box, the test suite has the following two failures because of this:

======================================================================
FAIL: testrepository.tests.commands.test_run.TestCommand.test_IDFILE_failures
----------------------------------------------------------------------
Text attachment: traceback
------------
Traceback (most recent call last):
  ...
  File "...\testrepository\tests\commands\test_run.py", line 100, in test_IDFILE_failures
    ], ui.outputs)
  ...
AssertionError: Match failed. Matchee: ...
Matcher: Equals(...)
Difference: !=:
reference = [('values',
  [('running', 'foo --load-list b:\\temp\\tmpsernri\\failing.list')]),
 ('popen',
  ('foo --load-list b:\\temp\\tmpsernri\\failing.list',),
  {'shell': True, 'stdout': -1, 'stdin': -1}),
 ('results', *),
 ('values', [('id', 1), ('tests', 0)])]
actual = [('values', [('running', 'foo --load-list b:\temp\tmpsernri\x0cailing.list')]),
 ('popen',
  ('foo --load-list b:\temp\tmpsernri\x0cailing.list',),
  {'shell': True, 'stdin': -1, 'stdout': -1}),
 ('results', <testrepository.ui.model.TestSuiteModel object at 0x00E27350>),
 ('values', [('id', 1), ('tests', 0)])]

------------
======================================================================
FAIL: testrepository.tests.test_testcommand.TestTestCommand.test_get_run_command_IDFILE_variable_setting
----------------------------------------------------------------------
Text attachment: traceback
------------
Traceback (most recent call last):
  ...
  File "...\testrepository\tests\test_testcommand.py", line 121, in test_get_run_command_IDFILE_variab
le_setting
    self.assertEqual(expected_cmd, fixture.cmd)
  ...
AssertionError: Match failed. Matchee: "foo --load-list b: emp mp_s0nwu"
Matcher: Equals('foo --load-list b:\\temp\\tmp_s0nwu')
Difference: !=:
reference = 'foo --load-list b:\\temp\\tmp_s0nwu'
actual = 'foo --load-list b:\temp\tmp_s0nwu'

------------

There is a wider problem with treating string configuration as shell script, in particular spaces other significant characters are passed through unescaped, which may also cause these tests to fail under some setups.