Add --build-tool-options to pass options to make/ninja/etc. (#25)

7688741
stopped us always using make but in the process removed the "-k" flag.

This flag causes make to carry on even when some targets fail. This is
very important for running the test suite in CI. Without it, a single
build failure marks thousands of subsequent programs as missing, as if
they too had failed to compile.

If I deliberately break a single source test the results currently are:
```
Total Discovered Tests: 3424
  Passed            :  378 (11.04%)
  Executable Missing: 3046 (88.96%)
Import succeeded.
```
This is what they should be:
```
Executable Missing Tests (1):
  test-suite :: SingleSource/UnitTests/2003-04-22-Switch.test

Total Discovered Tests: 3424
  Passed            : 3423 (99.97%)
  Executable Missing:    1 (0.03%)
Import succeeded.
```

cmake --build does not have its own -k option, so we have to pass it
after -- to the native tool. Unfortunately ninja's -k takes a number,
so ninja -k 0 is equivalent to make -k. Therefore we can't just
always add "-- -k" here.

Instead I've added a new option --build-tool-options where you can
pass any arguments you want. The build bots will use this to pass
"-k" to make, as we know for sure they use make.

Anything using ninja will also want to pass "-k 0" to get the same
effect.

Author: DavidSpickett

docs/quickstart.rst

@@ -72,6 +72,13 @@ command. The information below should be enough to get you started, but see the
    The ``SANDBOX`` value is a path to where the test suite build products and
    results will be stored (inside a timestamped directory, by default).
 
+   We recommend adding ``--build-tool-options "-k"`` (if you are using ``make``)
+   or ``--build-tool-options "-k 0"`` (if you are using ``ninja``). This ensures
+   that the build tool carries on building even if there is a compilation
+   failure in one of the tests. Without these options, every test after the
+   compilation failure will not be compiled and will be reported as a missing
+   executable.
+
 #. On most systems, the execution time results will be a bit noisy. There are
    a range of things you can do to reduce noisiness (with LNT runtest test-suite
    command line options when available between brackets):

lnt/tests/test_suite.py

@@ -580,7 +580,8 @@ def _build(self, path):
                               '--build', '.',
                               '-t', target,
                               '-j', str(self._build_threads())] +
-                             ([] if self.opts.succinct else ["-v"]),
+                             ([] if self.opts.succinct else ["-v"]) +
+                             ["--"] + shlex.split(self.opts.build_tool_options),
                              cwd=subdir)
         except subprocess.CalledProcessError:
             # cmake is expected to exit with code 1 if there was any build
@@ -1168,6 +1169,9 @@ def diagnose(self):
 @click.option("--use-make", "make", metavar="PATH",
               type=click.UNPROCESSED,
               help="Path to the build system tool [make/ninja/...]")
+@click.option("--build-tool-options",
+              help="Options to pass to the build system tool",
+              type=click.UNPROCESSED)
 @click.option("--use-lit", "lit", metavar="PATH", type=click.UNPROCESSED,
               default="llvm-lit",
               help="Path to the LIT test runner [llvm-lit]")