…w ones (#5401)

### Brief description of Pull Request

Adjust the ordering for component scheduling so that we give components
time to shutdown before we start new components. If components take too
long we will still move on but always wait for all components to shutdown
during scheduling.

### Pull Request Details

We have run in to issues with components which bind shared resources not
being able to release those resources when the component is rescheduled
due to a module rename / component rename. An example of this is the
`loki.source.syslog` component which will bind 1 or more ports depending
on the listener config. This causes issues with our current
implementation where shutdown/startup order is not defined,

1. Old components are told to shutdown
2. As soon as all old components have been told to shutdown, new
components are started
3. If the syslog component did not shutdown fast enough the new
component will fail to bind ports
4. Failing to bind ports does not cause the syslog component to exit we
just won't open the listener forcing a user to restart alloy to resolve
the problem

### Issue(s) fixed by this Pull Request

I need to look to see if there are issues which describe this sort of
behavior

### Notes to the Reviewer

This PR is intended to work with components which bind resources when
Run. We have multiple components which bind resources when the is newly
created before running. This causes a panic instead which will be
resolved in a separate PR.

I chose to leave an escape hatch on shutdown because our default
component shutdown timeout is 10 minutes. This is incredibly long to
leave a pipeline offline. Using the warning log window of 1 minute felt
much safer.

### PR Checklist

- [x] Tests updated

(cherry picked from commit 0fb959d)

…w ones [backport] (#5443)

## Backport of #5401

This PR backports #5401 to release/v1.13.

### Original PR Author
@kgeckhart

### Description
### Brief description of Pull Request

Adjust the ordering for component scheduling so that we give components
time to shutdown before we start new components. If components take too
long we will still move on but wait for all to shutdown during
scheduling.

### Pull Request Details

We have run in to issues with components which bind shared resources not
being able to release those resources when the component is rescheduled
due to a module rename / component rename. An example of this is the
`loki.source.syslog` component which will bind 1 or more ports depending
on the listener config. This causes issues with our current
implementation where shutdown/startup order is not defined,

1. Old components are told to shutdown
2. As soon as all old components have been told to shutdown, new
components are started
3. If the syslog component did not shutdown fast enough the new
component will fail to bind ports
4. Failing to bind ports does not cause the syslog component to exit we
just won't open the listener forcing a user to restart alloy to resolve
the problem

### Issue(s) fixed by this Pull Request

I need to look to see if there are issues which describe this sort of
behavior

### Notes to the Reviewer

This PR is intended to work with components which bind resources when
Run. We have multiple components which bind resources when the is newly
created before running. This causes a panic instead which will be
resolved in a separate PR.

I chose to leave an escape hatch on shutdown because our default
component shutdown timeout is 10 minutes. This is incredibly long to
leave a pipeline offline. Using the warning log window of 1 minute felt
much safer.

### PR Checklist

- [x] Tests updated


---
*This backport was created automatically.*

Co-authored-by: Kyle Eckhart <kgeckhart@users.noreply.github.com>