Improve `clash-testsuite` CLI interface

It would nice if the CLI for `clash-testsuite` was improved to

* allow easy control over which simulators are used to run the test suite
* provide more effective error reporting when simulators are not in `PATH`

For the first point, I propose adding `--simulator` and `--no-simulator` flags for each simulator we may use in the test suite. These would allow filtering of the tests based on simulator use in a convenient manner. For example

```
$ clash-testsuite --iverilog
```

would only run tests on `verilog`, and only with the `iverilog` simulator.

```
clash-testsuite --installed --no-verilator
```

would run on every simulator on the user's system with the exception of `verilator` (if it is installed). This means where we support multiple simulators for a particular HDL we can limit the jobs to the simulators we care about. The special extra flags `--all` and `--installed` could be used to force all simulators in the test suite / every simulator installed on a user's system.

Regarding reporting: currently when a simulator is not installed and it's tests are executed each test fails with a command not found error in its `stderr`. If we check in advance whether the necessary simulators are installed, we can error early without needing to generate code with `clash`, and provide a single useful error instead of generating HDL then failing for each test, e.g.

```
$ clash-testsuite --iverilog
ERROR: `iverilog` could not be found in PATH.
Please install `iverilog` or choose another simulator.
$ echo $?
127
```

Start at the clash-testsuite CLI entry point and trace how simulators are selected, checked, and invoked. Define completion as supporting per-simulator inclusion and exclusion flags plus --all and --installed, with an early, useful PATH error and exit status 127 instead of repeated test failures.

Written by the indexing model from the issue text.