fix: introduce null type to support resolver null expression

[HerrCai0907]

fix: #2662

The idea is compiler should infer the left type according to right type before compiling it.

[dcodeIO]

A more comprehensive fix for such cases could be, now that we have AnyClass <: Object, to compile null as Object instead of usize if there isn't a more specific context. Haven't tried, but if this works we'd also get rid of the "compiles to unusual type" warning.

[HerrCai0907]

@dcodeIO It is a break change, see this test code.
Of course, maybe we can do it, I do not think it is a good idea to make null as <usize>0. it just likes nullptr in cpp and NULL in c.

{ // only nulls infers as usize[]
  let arr1 = [null];
  assert(isInteger(arr1[0]));
  assert(!isNullable(arr1[0]));
  let arr2 = [null, null];
  assert(isInteger(arr2[0]));
  assert(!isNullable(arr2[0]));
}
{ // null in integer contexts infers as usize
  let arr1 = [1, null];
  assert(isInteger(arr1[0]));
  assert(!isNullable(arr1[0]));
  let arr2 = [null, 1];
  assert(isInteger(arr2[0]));
  assert(!isNullable(arr2[0]));
}

[HerrCai0907]

After trying it, null should not be ~lib/object/Object, consider this case, [null, "str"] should be inferred as Array<string | null>, but [nullable object, "str"] should be inferred as Array<object | null>

[CountBleck]

I can't say anything meaningful about big changes to the type system.