Debug radixcache: refactor recursive helper methods

[luzengxiangcn]

Motivation

Currently radixcache is implemented by several recursive functions, which may cause PYTHON Stack overflow problem when tree hits a certain depth. You may reproduce this issue by following code.

from openai import OpenAI
client = OpenAI(base_url="<sglang server url>", api_key="tests")
max_stack_depth = 5000
for i in range(max_stack_depth+10):
    print(i)
    messages = [{"role":"user", "content": "hello！" + "".join([" i"] * i)}]
    print(client.chat.completions.create(model="model", messages=messages, max_tokens=10))

Related Issue: #813
#1499 fixed one of recursive function but NOT ALL: .

Modifications

Replace recursion functions with loop or stack in radix_cache.py

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling.

[merrymercy]

cc @xiezhq-hermann @hnyls2002 Can you help review this?

[xiezhq-hermann]

Thank you for the contribution, can you make the change minimal and to the point? For example, put the style-wise change in a separated PR if needed.

[luzengxiangcn]

Thank you for the contribution, can you make the change minimal and to the point? For example, put the style-wise change in a separated PR if needed.

@xiezhq-hermann Sorry for the inconvenience caused to you due to the extensive code refactoring. To minimize the impact, I made minimal changes and pushed a new commit. I believe this will serve as a better starting point for the review.

Additionally, we can enhance the elegance of the code in the following two aspects， which can be done in another PR:

1. The logic of insert is essentially match logic + append new node, and the redundant parts can be merged to reduce code duplication.
2. The original helper functions were designed for iterative logic. After removing the iterative logic, the logic of these functions can be directly integrated into the original function.

[luzengxiangcn]

clean code and debug logic error

[luzengxiangcn]

@xiezhq-hermann Would you please help me review this? I fixed some logic errors in this PR, the code should be good now.