Create KeyItem struct for send_keys

[PotatoCP]

Modify behaviour of send_keys:

• Separates dispatch_typable and clear function from KeyInputState
• Return KeyItem instead of KeyboardEvent so we get can raw char and can convert it to action objects for dispatch actions.

Note that on spec, we should not immediately dispatch the event. Rather we should went through dispatch action steps.

Fixes #78.

[madsmtm]

Thanks for the PR!

• Separates dispatch_typable and clear function from KeyInputState
• Return KeyItem instead of KeyboardEvent so we get can raw char and can convert it to action objects for dispatch actions.

It would make it easier for me to review this if you could split it into two PRs?

Also, could you add a test that breaks with the old behaviour, but works with the new? That'd make it easier for me to understand the change.

[madsmtm]

Hmm, I'd like to avoid this breaking change. Why can you not use the character in KeyboardEvent::key's Key::Character(...) variant?

[PotatoCP]

My concern is that we can't get the raw character for Key::Named. Much smaller changes is maybe to simply add raw_char on the KeyboardEvent. Do you think we should do that instead?

This change does not exactly fix behaviour for this crate, but it makes it simpler for user like web engine to implement Webdriver Element Send Keys command based on spec.

[madsmtm]

Returning back to this:

I'm actually somewhat inclined to remove the webdriver module in v0.9.x, and instead split it into a different crate, to allow it to evolve at a different pace than the main keyboard-types crate, which otherwise contains (mostly) types that the entire ecosystem depends on for interop with other crates.