Nonexistent benchmark should give better error message

When trying to run a nonexistent benchmark, error message should say does not exist. Currently it says it's not valid on debian.

```
dlott@dlott:~/git/PerfKitBenchmarker$ ./pkb.py --benchmarks does_not_exist
2017-03-08 14:36:13,654 b539f88b MainThread INFO Verbose logging to: /tmp/perfkitbenchmarker/runs/b539f88b/pkb.log
2017-03-08 14:36:13,654 b539f88b MainThread INFO PerfKitBenchmarker version: v1.11.0
2017-03-08 14:36:13,655 b539f88b MainThread INFO Flag values:
--benchmarks=does_not_exist
Traceback (most recent call last):
File "./pkb.py", line 21, in
sys.exit(Main())
File "/usr/local/google/home/dlott/git/PerfKitBenchmarker/perfkitbenchmarker/pkb.py", line 726, in Main
return RunBenchmarks()
File "/usr/local/google/home/dlott/git/PerfKitBenchmarker/perfkitbenchmarker/pkb.py", line 650, in RunBenchmarks
benchmark_specs = _CreateBenchmarkSpecs()
File "/usr/local/google/home/dlott/git/PerfKitBenchmarker/perfkitbenchmarker/pkb.py", line 327, in _CreateBenchmarkSpecs
benchmark_tuple_list = benchmark_sets.GetBenchmarksFromFlags()
File "/usr/local/google/home/dlott/git/PerfKitBenchmarker/perfkitbenchmarker/benchmark_sets.py", line 227, in GetBenchmarksFromFlags
(benchmark_name, FLAGS.os_type))
ValueError: Benchmark "does_not_exist" not valid on os_type "debian"
```