Add Kumru-based reasoning tokenizer with fixed system instruction template

Delete rslm-byte-bpe-vocab.json

Delete tokenizer_training_report.json

Delete tokenizer_stats.json

Delete special_tokens_map.json

Delete rslm-byte-bpe-merges.txt

Delete TOKENIZER_TODO.json

Add Kumru-based reasoning tokenizer with fixed system instruction template

Add CPU-safe 65K RSLM tokenizer trained on BellaTurca

Add CPU-safe 65K RSLM tokenizer trained on BellaTurca

Add RSLM tokenizer trained on AkademikDerlem/makaleler

Add RSLM-1B-Speed architecture and weights

initial commit