bthread可能突破数量限制

**Describe the bug (描述bug)**
现在每个worker的bthread数量限制是4096，但是bthread_start_urgent的产生不会校验worker上面的bthread的数量，这样一个worker上面的bthread数量就有可能突破4096。这块是否应该在bthread_start_urgent加个校验？这块加校验感觉会影响性能。
或者在EventDispatcher::Run()里面做一些校验？

**To Reproduce (复现方法)**

**Expected behavior (期望行为)**

**Versions (各种版本)**
OS:
Compiler:
brpc:
protobuf:

**Additional context/screenshots (更多上下文/截图)**

Start by reading the bthread_start_urgent path and EventDispatcher::Run() to understand where the per-worker limit of 4096 is enforced or bypassed. Compare the performance implications of checking in each location; done means preventing the worker’s bthread count from exceeding the limit without an unacceptable performance impact.

Written by the indexing model from the issue text.