refactor: style recommendation and new recommended severity

[ematipico]

Summary

Part of #4662

style rules aren't recommended anymore

It's possible that people won't agree with this choice, so let's discuss it. However, our recommended rules are a lot, and we should slim them down.

With the introduction of domains, there's also a faster and better (granular) way to add recommended rules.

Styling rules can be very opinionated, and we should avoid imposing a tool's opinion on users unless there's a design behind it. Our linter wasn't born to be opinionated by design, as opposed to our formatter. In the last months, I could see some displeasure among our users, asking why certain style rules were recommended, considering that until v1.*, the recommended ones were emitting an error.

Important

We can still revert the change, however I would like reduce the number of errors and noice. For example, noVar shouldn't be an error.

New "on" option

Until v1, when users wanted to enable non-recommended rules, they needed to choose a severity from info, warn, and error.

However, now that all rules have a default severity, users can decide to use the severity suggested by Biome. In the following example, when a user uses "on" for noArguments, violations of this rule will raise errors, because that's its default severity suggested by us. Instead, noNamespace will raise an info diagnostics:

{
	"linter": {
		"rules": {
			"style": {
				"noArguments": "on",
				"noNamespace": "on"
			}
		}
	}
}

Better biome explain <RULE>

I didn't like the explanation of our rules. I revamped it, here's a preview from my terminal:

Test Plan

Added new tests and update many of our current tests

[codspeed-hq]

CodSpeed Performance Report

Merging #4730 will not alter performance

_{Comparing feat/on-rules (954547f) with next (0a9d85a)}

Summary

✅ 97 untouched benchmarks

[Conaclos]

I like this change! I suggest using Default severirity instead of Suggested severity in biome explain. This is less confusing IMO.

[arendjr]

Didn’t have time for a full review, but I’m on board too 👍

Just two suggestions:

• I think biome migrate should keep them enabled for users coming from Biome 1.x. For new users we can indeed leave them disabled by default going forward.
• For noVar specifically: var actually has confusing semantics, and can cause bugs in combination with loops, for instance. It feels this more than just a style choice, so maybe we need to recategorize the rule?

[ematipico]

so maybe we need to recategorize the rule?

Definitely wise to do that. Maybe we can move it to the suspicious group

I think biome migrate should keep them enabled for users coming from Biome 1.x. For new users we can indeed leave them disabled by default going forward.

That's wise, I will follow up with another PR