feat(rome_js_analyze): noInferrableTypes

[Conaclos]

Summary

Closes #4084

This implements ESLint's no-inferrable-types with some change:

• property parameters are supported
• readonly properties are supported
• we accept wide types in const contexts. This allows to hide information: this is particularly useful when exporting a constant (export const C: number = 1). This also looks more consistent (ESLint allows const x: 1 | 2 = 1. Thus why const x: number = 1 should be rejected?)
• primitive type constructors are not supported (String(1), Number('1') ...):. They may be shadowed
• undefined value is not supported (undefined may be shadowed)
• type RegExp is not supported because it may be shadowed
• we do not support the options that allow disabling the rule for default parameters and/or properties.

Test Plan

Includes a large set of unit tests.

Documentation

• The PR requires documentation
• I will create a new PR to update the documentation

[netlify]

✅ Deploy Preview for docs-rometools canceled.

Built without sensitive environment variables

Name	Link
🔨 Latest commit	ff13139
🔍 Latest deploy log	https://app.netlify.com/sites/docs-rometools/deploys/63e9753974bdf60008bedfad

[ematipico]

The fix of the rule can't be flagged as "safe" because it removes trivia, as we can see in this snapshot. If removing trivia is the intended behaviour, then the fix should be Applicability:MaybIncorrect

[Conaclos]

This was not interventional. I fixed the issue.

[ematipico]

Looks good, we can merge once rebased and conflicts are resolved

[Conaclos]

Looks good, we can merge once rebased and conflicts are resolved

Done :)