BTRIM function, octal and binary literals and other changes

[katzyn]

1. Parser is now able to parse multiple PostgreSQL-style casts (expression::dataType1::dataType2) and multiple array element references (array[index1][index2]).
2. Comparison operations between different numeric data types don't create intermediate values any more. Casts between numeric data types and character strings are now able to handle values larger than limits of these numeric data types, closes In comparison text values are converted to INT even when they should be converted to BIGINT #352.
3. GREATEST, LEAST, LPAD and RPAD functions are now part of the SQL standard, vendor-specific mark from documentation is removed. Implementations of LPAD and RPAD now can use non-BMP characters for padding, but length still needs to be specified in UTF-16 characters.
4. LTRIM, RTRIM, and BTRIM functions are now part of the SQL Standard and they support multiple trimmed characters in the Standard. Implementations of LTRIM and RTRIM are enhanced to support multiple characters, a new BTRIM function is introduced. Implementations of all these functions and TRIM function are also improved to support non-BMP characters.
5. Standard now has hexadecimal, octal, and binary literals for integer values. Hexadecimal literals were already supported and tokenizer is improved to support octal and binary literals.