[nit] Make sure Ray Serve installs ray[default]

[triciasfu]

No description provided.

[edoakes]

For this, let's add a test that:

1. Creates a fresh conda env
2. Installs ray[serve] only (ideally it'd be from master, maybe not feasible?)
3. Checks the output of a simple e2e serve script

@simon-mo @architkulkarni could we re-use the runtime_env testing mechanism for installing the wheel built in CI for (2)?

[edoakes]

I tried running this:

import ray
from ray import serve
serve.start()

@serve.deployment(ray_actor_options={"runtime_env": {"pip": ["requests"]}})
class A:
    pass

A.deploy()

and it hangs without any warning:

(pid=80782) 2021-10-14 09:39:56,527     INFO backend_state.py:896 -- Adding 1 replicas to deployment 'A'. component=serve deployment=A
2021-10-14 09:40:14,882 WARNING worker.py:1227 -- The actor or task with ID ffffffffffffffff0c267d7c776f95260c6091e201000000 cannot be schedule
d right now. You can ignore this message if this Ray cluster is expected to auto-scale or if you specified a runtime_env for this actor or task
, which may take time to install.  Otherwise, this is likely due to all cluster resources being claimed by actors. To resolve the issue, consid
er creating fewer actors or increasing the resources available to this Ray cluster.
Required resources for this actor or task: {CPU_group_7c45bd96450f2713ebbdfbc054264143: 1.000000}
Available resources on this node: {7.000000/8.000000 CPU, 274103400.000000 GiB/274103400.000000 GiB memory, 137051700.000000 GiB/137051700.0000
00 GiB object_store_memory, 1000.000000/1000.000000 bundle_group_7c45bd96450f2713ebbdfbc054264143, 0.000000/1.000000 CPU_group_7c45bd96450f2713ebbdfbc054264143, 1.000000/1.000000 CPU_group_0_7c45bd96450f2713ebbdfbc054264143, 0.980000/1.000000 node:172.20.6.33, 1000.000000/1000.000000 bundle_group_0_7c45bd96450f2713ebbdfbc054264143}
In total there are 0 pending tasks and 1 pending actors on this node.
(pid=80782) 2021-10-14 09:40:26,614     WARNING backend_state.py:1072 -- Deployment 'A' has 1 replicas that have taken more than 30s to start up. This may be caused by waiting for the cluster to auto-scale or because the constructor is slow. Resources required for each replica: {'CPU': 1}, resources available: {'CPU': 7.0}. component=serve deployment=A
(pid=80782) 2021-10-14 09:40:56,690     WARNING backend_state.py:1072 -- Deployment 'A' has 1 replicas that have taken more than 30s to start up. This may be caused by waiting for the cluster to auto-scale or because the constructor is slow. Resources required for each replica: {'CPU': 1}, resources available: {'CPU': 7.0}. component=serve deployment=A
(pid=80782) 2021-10-14 09:41:26,741     WARNING backend_state.py:1072 -- Deployment 'A' has 1 replicas that have taken more than 30s to start u
p. This may be caused by waiting for the cluster to auto-scale or because the constructor is slow. Resources required for each replica: {'CPU':
 1}, resources available: {'CPU': 7.0}. component=serve deployment=A
(pid=80782) 2021-10-14 09:41:56,813     WARNING backend_state.py:1072 -- Deployment 'A' has 1 replicas that have taken more than 30s to start u
p. This may be caused by waiting for the cluster to auto-scale or because the constructor is slow. Resources required for each replica: {'CPU':
 1}, resources available: {'CPU': 7.0}. component=serve deployment=A
(pid=80782) 2021-10-14 09:42:26,910     WARNING backend_state.py:1072 -- Deployment 'A' has 1 replicas that have taken more than 30s to start u
p. This may be caused by waiting for the cluster to auto-scale or because the constructor is slow. Resources required for each replica: {'CPU':
 1}, resources available: {'CPU': 7.0}. component=serve deployment=A

[simon-mo]

hint: just duplicate the ray minimal install tests and add some smart things to runtime env (serve inject ray serve into runtime env)