Wrong offset documentation for calendar:system_time_to_rfc3339/2

[eproxus]

Describe the bug
The type offset() is defined as:

[byte()] | (Time :: integer()).

The documentation for calendar:system_time_to_rfc3339/2 states:

{offset, Offset} - The offset, either a string or an integer, to be included in the formatted string. An empty string, which is the default, is interpreted as local time. A non-empty string is included as is. The time unit of the integer is the same as the one of Time.

(my emphasis)

However, only the value "Z" is allowed.

To Reproduce

1> calendar:system_time_to_rfc3339(erlang:system_time(), [{unit, nanosecond}, {offset, "Foobar"}]).
** exception error: bad argument
     in function  list_to_integer/1
        called as list_to_integer("oo")
        *** argument 1: not a textual representation of an integer
     in call from calendar:offset_string_adjustment/3 (calendar.erl, line 975)
     in call from calendar:system_time_to_rfc3339_do/4 (calendar.erl, line 698)
2> calendar:system_time_to_rfc3339(erlang:system_time(), [{unit, nanosecond}, {offset, "A"}]).
** exception error: no match of right hand side value "A"
     in function  calendar:offset_string_adjustment/3 (calendar.erl, line 974)
     in call from calendar:system_time_to_rfc3339_do/4 (calendar.erl, line 698)

Expected behavior
Either the documentation (and type) should say only "Z" is allowed, or the function should allow any string as documented.

Affected versions
27.1.2

Additional context
The RFC's grammar only allows Z as a letter, so the documentation and type should probably be fixed instead. If that is the desired outcome, I can make a PR for that.

[garazdawi]

Hello! Yes, the documentation should be updated to only specify "Z" / "z" / "+HH:MM" / "-HH:MM". I don't think we can make the type any more precise as the closest we get is [$Z,$z,$+,$-,$0,$1,$2,$3,$4,$5,$6,$7,$8,$9,$:] which is not a lot more helpful versus the current type.