Raise delay for first-request-exceeds-server-limit test

[anuraaga]

In connect-python, we've found this one test to flake fairly commonly on CI

https://github.com/connectrpc/connect-python/blob/0efb7b06693c9f4ce53bb45d0c5f357fb85a4947/conformance/test/test_server.py#L33

The hypothesis is that handling the first message takes longer than any other message since it also sets up the request machinery itself and can take some time on slow CI runs.

It's challenging to push a custom binary to CI to verify the effect but wonder if it makes sense anyways to make the change. My understanding is

• The delay is a cap on test time, where a failing test will reach it, and a non-failing one won't be affected
• subsequent-request-exceeds-server-limit sends too messages with 50ms delay for total of 100ms so this allows the delays to match. Note that we don't see any flakiness with subsequent-request-exceeds-server-limit

/cc @stefanvanburen