chore: Set up semantic release

[dcodeIO]

As discussed on Slack and in #1331, this PR removes our custom release logic, including publishing nightlies which isn't necessary anymore after the change, and installs semantic-release (currently configured to perform dry runs in case something goes wrong).

Commit format would then be:

• fix: Message for a patch (wondering what will happen there with version < 1)
• feat: Message for a minor
• BREAKING CHANGE: Message (or used as a footer) for a major

• I've read the contributing guidelines

[dcodeIO]

Looking up the v0 problem revealed that they are apparently not handling v0.x.x as one would expect, but will bump to v1.0.0 on the next breaking change. Quite unfortunate.

[MaxGraey]

Btw it seems it could be with optional scope like:
fix(parser): message
feat(std): message
In this case will be great have whole list of conventions

[dcodeIO]

Yeah, using a scope seems good. I'm however unwilling to force out a v1 just because of this. Now one could argue that version numbers are arbitrary anyway, but I guess if there is a case to be made about something being "not quite ready" to be a v1.0.0, then it's in tech like WebAssembly, so I'd like to stick to v0.x.x for now.

[dcodeIO]

Now uses my fork of semantic-release that has a --unstable command line option enforcing releases to stay in the v0.x.y range. Turned out this can be achieved with just a few lines of code.

[dcodeIO]

Merging, but letting it perform dry runs initially so I can check the logs, hence no actual releases just yet. Going to activate once sufficiently confident that it's working :)