And skip testwsdeque if it is not in-tree, as we rely on some headers
from the build tree.

This increases the parallelism possible, and allows us to track what
progress we are making.

I think different Linux setups generate different output for these tests,
so we just ignore the output on Linux now.

Also, clean up the way we find the output file. From the comment:

# Finding the sample output.  The filename is of the form
#
#   <test>.stdout[-<compiler>][-<version>][-<wordsize>][-<platform>]
#
# and we pick the most specific version available.  The <version> is
# the major version of the compiler (e.g. 6.8.2 would be "6.8").  For
# more fine-grained control use if_compiler_lt().

I'll update the wiki too.

Now instead of saying, e.g.
    namebase_if_compiler_lt('ghc','6.9', 'foo-6.8')
you say
    if_compiler_lt('ghc','6.9', namebase('foo-6.8'))

Allows us to use a different name base for foo.stderr for old compilers, e.g.
test('tc141',
     namebase_if_compiler_lt('ghc', '6.9', 'tc141-6.8'),
     compile_fail,
     [''])

This is better than a top-level clean() when using threads

The user001 test in the unix package can fail when stdin comes from a
file.

When working on a new foo extension, you can now put your tests in the
testsuite, set ProjectTags=-foo in mk/build.mk and skip_unless_tag('foo')
in the tests.

This means we can put package-specific tests in the repository for the
package, rather than putting them in the testsuite.  There should be a
.T file to go with the tests, in the same way as for other tests in
the testsuite (but this could be in addition to a standalone test
driver that works with Cabal's 'setup test').

I had to pull the global classes and instances out into their own module
as there was a catch-22: testlib needed to know if threading was enabled,
but we don't know that until we have gone through the argument, but going
through the arguments required changing things like config in testlib.