Docs: [no-unnecessary-condition] mention subtleties of primitives' assignability

[kirkwaiblinger]

Before You File a Documentation Request Please Confirm You Have Done The Following...

• I have looked for existing open or closed documentation requests that match my proposal.
• I have read the FAQ and my problem is not listed.

Suggested Changes

I'd like to add a section that briefly alerts users to the nuances around primitives being assignable to object types, causing sort-of-false-positives with no-unnecessary-condition's truthiness checks.

I'd like to mention:

• "object types" are just "things with properties", not objects (a superset of no-empty-object-type shenanigans)

  function foo(x: { toString: Function }) {
     if (x) { // reported by no-unnecessary-condition as always-truthy
     }
  }
  
  foo(false); 
  foo(true);
  foo({});

• object can easily contain primitives at runtime, without typical obviously unsafe TS patterns.

  function foo(o: object) {
    if (o) { // reported by no-unnecessary-condition as always-truthy
    }
  }
  
  foo(false); // TS error.
  
  const lol: {} = false;
  foo(lol); // but this is allowed by TS.

As said, I think this should be brief, just to alert users to possibly-faulty common assumptions about the type system, and then punt deeper dive to microsoft/TypeScript#60582 (comment) or any other good references. The main goal is really to convey think closely about your application before "fixing" a truthiness report, since the condition may actually be necessary, despite what the types appear to be indicate.

Affected URL(s)

https://main--typescript-eslint.netlify.app/rules/no-unnecessary-condition#limitations

Additional Info

Relates to #10378, which may eventually change some of this behavior. But regardless of rule behavior change on our side, it's valuable for users to know that primitives may be lurking places they didn't expect.