Improve SPACE(n) emulation where possible

[lukaseder]

As can be visualised with our new manual sections for different functions, and their emulations:
https://www.jooq.org/doc/3.14/manual/sql-building/column-expressions/string-functions/space-function/

The SPACE(n) emulation is more complicated than it needs to be in many dialects:

-- FIREBIRD
cast(rpad(' ', (char_length(' ') * 3), ' ') as varchar(4000))

-- CUBRID, H2, MARIADB, MYSQL, ASE, AURORA_MYSQL, DB2, SQLSERVER, SQLDATAWAREHOUSE, SYBASE, VERTICA
space(3)

-- INFORMIX, MEMSQL
rpad(' ', (char_length(' ') * 3), ' ')

-- HANA, INGRES, ORACLE, TERADATA
rpad(' ', (length(' ') * 3), ' ')

-- SQLITE
"replace"(hex(zeroblob(3)), '00', ' ')

-- HSQLDB, POSTGRES, AURORA_POSTGRES, COCKROACHDB, REDSHIFT
repeat(' ', 3)

It uses repeat() internally, but with a literal(' '), whose length is known by jOOQ, so the length doesn't need to be calculated dynamically.

---

See also: https://twitter.com/Avalanche1979/status/1255089041005056001