Add chunking for even better generated git messages

[eonist]

1. pre-rinse diff payload
2. smartly chunk the total payload say 100kb
3. 4k context window. is about 16kb so 10kb for input equals 10 step for loop
4. call afm llm model 10 times and get special summary
5. combine summaries into 1 unified summary

considerations:

• if diff size is extremely big, we need to have some strategies to smartly processes things. because even the interim partial summaries will be too large for the final output summarization step etc. open for suggestions here etc

[eonist]

Solution for Chunking Large Diffs-2.md
Solution for Chunking Large Diffs.md
Hierarchical Chunking Strategy for Large Diffs.md
Solution for Chunking Large Git Diffs.md