Improve the Omit helper type

[vipcxj]

Fixes #31153
Since Omit is the offical helper type now, I think it should be a perfect version. The common version not work with type like { a: string, b: number, [key: string]: any, }, this version works. The detail see #31153

[vipcxj]

A few test failed. But I think it is the result of other bug.

function test<T extends { a: string, b:string }>(obj: T): T {
  let { a, ...rest } = obj;
  const test = rest as Omit<T, 'a'>;
  test.b // it's ok;
  const test2 = { a, ...test };
  console.log(test2.a) // it's ok;
  console.log(test2.b) // it's ok;
  test2.a = '1' // it's ok;
  test2.a = 1 // it's not ok, test2.a must be string;
  test2.b = '1' // it's ok
  test2.b = 1 // it's not ok, test2.b must be string;
  test2.c = 1 // it's not ok, c does not exist on type `type of test2`;
  // so it seems that the type of test2 should be { a: string, b:string }
  const test3 = test2 as T // not ok, TS2352, why? Is there a tool to see what's the real type of { a: string } & Omit<T, 'a'>
}

[dragomirtitian]

The PR template mentions that you can submit fixes for issues in the backlog I believe (the issue you are referencing is a closed question)

Also I don't think adding this more complicated Omit is necessarily a good idea, since as I mention in the comments to 31153 this may need to be changed once we have generalized index signatures form #26797.

[vipcxj]

@dragomirtitian This is not a problem if the generalized index signatures form is properly restricted. I think the index signatures without restricted will be a disaster.

[RyanCavanaugh]

Touching Omit in any fashion is going to be a massive breaking change. It is also intentional that Omit does not constrain its second argument.

Please only submit PRs for issues in the Backlog milestone.