Add benchmarking tool

[bruno-f-cruz]

This PR adds a new tool to run hardware tests of Harp devices against the Harp protocol specification.

The main features:

• Integrates with the toolkit using the benchmark command: dotnet run --project Harp.Toolkit benchmark --port COM3 --report here.html
• Takes as single required input the port of the device to test
• Outputs output to the console using ANSI (also supports verbose mode to render all test results via a --verbose flag)
• Using RazorLight templates, can also render the results to a html file that can be used to exchange results

[bruno-f-cruz]

@glopesdev was this the integration you had in mind for new tools?
Mostly ok with it. Just a few questions:

• Should we create new namespaces to not clutter the top level one (e.g.: Harp.Toolkit.Benchmark) or leave everything as Harp.Tookit
• Are you ok with adding these two extra dependencies? (RazorLight and Ansi support)

[glopesdev]

Should we create new namespaces to not clutter the top level one (e.g.: Harp.Toolkit.Benchmark) or leave everything as Harp.Tookit

@bruno-f-cruz Happy to create a namespace for particularly complex commands requiring many helper classes.

An alternative is to create a separate library for the command backend and then add it as a dependency (e.g. I am going to do this with the reflex generator).

Are you ok with adding these two extra dependencies? (RazorLight and Ansi support)

In principle not opposed to any dependencies, since this will be a self-contained dotnet tool, so it will be deployed and run completely independent from any bonsai project dependency, etc.

But let's discuss the details on the benchmarks and reflex generator PRs.

[bruno-f-cruz]

Sounds good. I guess the best way to go about this is to expose a single method (e.g.: https://github.com/harp-tech/Harp.Benchmark/blob/bb0529326aee7a12ab9b33d2dfeec573f8bbd2ac/src/Harp.Benchmark/Program.cs#L43) as the entry point of the tool and create a single command subclass in this repository that wraps it?

[bruno-f-cruz]

@glopesdev took a first stab at this that would benefit from initial comments. Feel free to run the tool against a device; it should output something sensible. We can discuss some edge cases during one of the upcoming harp meetings

[banchan86]

Decided to give this a try and I get random unauthorized port access errors on successive runs with just the core benchmark tests. It seems that each test opens its own connection, and the handle is not being released in time before the next test opens. Once it hits a snag a whole bunch of tests fail before it recovers. This might only surface on slower PCs, but curiously it does not happen when I run it the full device tests with the device.yml option and I'm not sure why 😆

HTML Reports	device-yml	Unauthorized errors
behavior, behavior2, behavior6, behavior7	no	10, 1, 42, 18
behavior4, behavior5, behavior8, behavior9	yes	0, 0, 0, 0

Also, if the device isn't connected and the com port doesn't exist, the report says:

"System.IO.FileNotFoundException: Could not find file 'COM9'." when something like the port could not be found might be a more obvious error message (and ideally it should throw that error without even running the tests). The behavior3.html report is from a run when I disconnected the device.

HTML reports attached!

behavior.zip