Cline 4.1.19 stops guessing when to compact

The compaction trigger had compared a character-based estimate of the conversation, around three characters per token, against the model's context limit. Content that tokenizes denser — disassembly, image dumps, minified sources — filled the real window while the estimate stayed under the threshold, compaction never fired, and the turn was squeezed into a handful of output tokens. The trigger now also uses the token count the provider reports.⁶