`array_position` doesn't check nulls in array index & fails to handle nulls properly

[Jefffrey]

Describe the bug

Test cases:

select array_position(haystack, needle, index_from)
from values
  ([1, 2], 1, 2),
  ([1, 2], 1, null)
t(haystack, needle, index_from)

• Currently this fails because 0 is out of bounds, but this is only because by default the value underneath a null is usually 0; we can construct an array where this isn't the case and pass in a valid in-bounds value even if it should be masked by a null
• This should instead fail by checking nulls, as scalar path also raises exec error if it detects a null for the index argument

> select array_position(haystack, needle), array_positions(haystack, needle)
from values
  ([[1], null], null),
  ([[]], null)
t(haystack, needle);
+-------------------------------------+--------------------------------------+
| array_position(t.haystack,t.needle) | array_positions(t.haystack,t.needle) |
+-------------------------------------+--------------------------------------+
| NULL                                | []                                   |
| 1                                   | [1]                                  |
+-------------------------------------+--------------------------------------+

• This should return 2 | [2] for the first row and NULL | [] for the second row
• This is because in the first row, we can find a null element so we return that position
• And for the second row we can't find any null, so we shouldn't return a position

To Reproduce

No response

Expected behavior

No response

Additional context

No response

[Jefffrey]

Related:

• Сustomize behavior of NULLS in array functions #6808