fix function signatures for osmAuth

[k-yle]

Follow up from openstreetmap/iD#11661 (comment)

This reverts #125, since that PR broke IDE-autocomplete and type-safety for anyone who uses x = osmAuth() rather than x = new osmAuth().

Declaring osmAuth as an ES6 class is wrong, since it's just a function. Therefore it's valid to instantiate the class without the new operator. The example in the README also does not use new.

This is not an exact revert of #125, instead it settles on a compromise that should work for everyone. Therefore, this is not a breaking change.

---

Tested by creating a new file in the repo called test.ts with the following content:

import { osmAuth } from 'osm-auth';

const a = osmAuth({
  client_id: '',
  scope: '',
  redirect_uri: '',
  // @ts-expect-error -- typo should be detected
  invalid: 1,
});
a.authenticate(console.log);
// @ts-expect-error -- typo should be detected
a.invalid();

const b = new osmAuth({
  client_id: '',
  scope: '',
  redirect_uri: '',
  // @ts-expect-error -- typo should be detected
  invalid: 1,
});
b.authenticate(console.log);
// @ts-expect-error -- typo should be detected
b.invalid();

If the file compiles with no errors, that proves that this PR works for both cases.

[bhousel]

We should probably just rewrite this thing to be an actual ES6 class.

[k-yle]

We should probably just rewrite this thing to be an actual ES6 class.

Would you accept a PR that does this? the diff would be massive (adding this. as a prefix to many functions/variables), so it might be hard to review.

maybe the best short-term option is just to update the README, to suggest that people use new? If you use new, the existing definitions works well

[bhousel]

Yeah for now let's just update the docs to encourage new, that seems like the simplest thing.

[k-yle]

cool, then I'll close this and we can follow up on the docs in #148