add btoa()/atob() builtins

- Introduce global btoa() and atob() functions
- Encoder: fast 12-bit pair-LUT, ~3.6 GB/s
- Decoder: branchless streaming form, ~0.65 GB/s scalar
- Tolerant to whitespace, validates padding and invalid input
- Minimal allocations: only one malloc if input is wide-char
- Fully compliant with DOMException

Co-authored-by: Saúl Ibarra Corretgé <s@saghul.net>

Author: Lucas