Support VARCHAR/VARBINARY without explicit length

[mrotteveel]

Currently, Firebird doesn't have a default length for VARCHAR/CHARACTER VARYING, { NCHAR | NATIONAL {CHAR|CHARACTER} } VARYING, and VARBINARY/BINARY VARYING.

This PR adds a default length of 255 for all these types.

Why 255:

• Big enough for common use cases
• Big enough to cast from other (non-string) scalar types without truncation
• Small enough to not be excessive in storage use, and fits in an index on page size 8192 with UTF8, with some room to spare (e.g. for compound indices)
• 0xFF ;)

Why at all:

• Required by the SQL standard (see below)
• Allows things like CAST(something as VARCHAR) (see also second point under "Why 255"), which can be a simplification for query generators
• Can result in simpler code for ad-hoc queries and such

For reference, the SQL:2023 standard requires such a default, as 6.1 <data type> defines:

<character string type> ::=
    [...]
  | CHARACTER VARYING [ <left paren> <character maximum length> <right paren> ]
  | CHAR VARYING [ <left paren> <character maximum length> <right paren> ]
  | VARCHAR [ <left paren> <character maximum length> <right paren> ]
  | [...]

[...]

<binary string type> ::=
    [..]
  | BINARY VARYING [ <left paren> <maximum length> <right paren> ]
  | VARBINARY [ <left paren> <maximum length> <right paren> ]
  | [...]

[...]

<national character string type> ::=
    [...]
  | NATIONAL CHARACTER VARYING [ <left paren> <character maximum length> <right paren> ]
  | NATIONAL CHAR VARYING [ <left paren> <character maximum length> <right paren> ]
  | NCHAR VARYING [ <left paren> <character maximum length> <right paren> ]
  | [...]

With syntax rule 6 (for BINARY) and 7 (for VARCHAR/NCHAR VARYING) specifying:

6) If <maximum length> is omitted, then an implementation-defined (IL007) <maximum length> is implicit.

7) If <character maximum length> is omitted, then an implementation-defined (ID069) <character maximum length> is implicit.

See also thread Introducing a default VARCHAR length on firebird-devel.

[aafemt]

Instead of doubling branches I would introduce opt_char_length clause that returns 255 on absence and replace with it '(' pos_short_integer ')' in existing code. Did you try it?

[mrotteveel]

Instead of doubling branches I would introduce opt_char_length clause that returns 255 on absence and replace with it '(' pos_short_integer ')' in existing code. Did you try it?

I also have to track FLD_has_len, which is needed for SET BIND (and maybe other things), and this follows the pattern that is also applied for BINARY, CHAR and NCHAR.

[mrotteveel]

@dyemanov Any objections if I merge? (The build failure is a transient failure that occurs regularly on macOS.)