Function worker config item "maxPendingAsyncRequests" not passed to function instances under Process or Kubernetes Runtimes

[csthomas1]

Describe the bug
When I enable either the Process or Kubernetes function Runtimes and set the parameter "maxPendingAsyncRequests" to a value other than 1000 in the function worker configuration (e.g. functions_worker.yml), any function instances launched by the worker continue to have maxPendingAsyncRequests = 1000. I'd like to be adjust this parameter to guard against my function running out of memory -- without increasing the function heap size -- due to too many publish operations in flight, as can occur when the output topic's backlog quota has been reached.

Investigation shows that the problem lies in class pulsar-functions/runtime/src/main/java/org/apache/pulsar/functions/runtime/RuntimeUtils.java, method getCmd, which simply does not include the necessary argument for maxPendingAsyncRequests from the InstanceConfig.

To Reproduce
Steps to reproduce the behavior:

1. Enable the Process Runtime in the broker:
   • Using the pulsar docker image, set the following environment variables when running the pulsar broker:
     • PF_functionRuntimeFactoryClassName: org.apache.pulsar.functions.runtime.process.ProcessRuntimeFactory
     • PF_containerFactory: process
2. Set the maxPendingAsyncRequests property to a value other than 1000
   - also in the pulsar broker container set the following environment variable:
   - PF_maxPendingAsyncRequests: "250"
3. Launch a pulsar function
4. Look at the top of the pulsar function log, find the dump of the InstanceConfig. You should see "maxPendingAsyncRequests=1000" following the settings for "maxBufferedTuples", "functionAuthenticationSpec", "port", and "clusterName".
   Expected behavior
   When performing the above steps, I expect to see "maxPendingAsyncRequests=250".

Additional context
(bugfix pull request incoming)

[zymap]

Move this to the next release.