isNaN usage not covered by tests

[johnd0e]

There are some (~5) occurrences of isNaN usage in code, and no single test to understand their real purpose.
And that function is known to be not so obvious: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/isNaN
It's not straightforward to make sure that value is number, because boolean and empty string also coercing to numbers.

So I am not sure that our current usage is correct.
And more - instead of coercing numeric strings - I'd prefer throwing errors

Thus, we need to refactor that parts and cover them by tests.

[johnd0e]

Ref: #7128 (comment)

[johnd0e]

1. isNaN itself is not reliable way to check if value can be converted to number. There a lot of discussions, and I just point to these unit tests: https://run.plnkr.co/plunks/93FPpacuIcXqqKMecLdk/
   So I propose to implement isNumeric like here: LatLng: stricter arguments validation #7765
2. Just making sure that the value is convertible to number is not enough. Instead we should actually convert the value to number as early as possible (otherwise '5'+5 = '55').
   There are some places in code vulnerable to this sort of bugs, and a number of open issues (too lazy to link them right now).

[itsyourOOP]

@johnd0e

Is the issue still open? Because if it is, I'd like to try if that's fine.

[Falke-Design]

@itsyourOOP I think this not a good issue to start, it is currently not clearly how we want to further with the number check in Leaflet.

A easy issue to start would be: #7787