Description of the problem:

Multicluster_operators_hub_subscription pod having various issues due to the constantly increasing /tmp storage usage.

This includes:
- Error: Kubelet may be retrying requests that are timing out in CRI-O due to system load: context deadline exceeded: error reserving ctr name k8s_multicluster-operators-hub-subscription_multicluster-operators-hub-subscription-..._open-cluster-management_... for id 3844...c6b9: name is reserved

- Pod constantly being evicted with error: 
The node was low on resource: ephemeral-storage. Container multicluster-operators-hub-subscription was using 16392Ki, which exceeds its request of 0.

- Pod logging errors similar to:
E0209 13:08:19.899823       1 gitrepo_sync.go:388] lstat /tmp/data-services/release/internal: no such file or directoryFailed to sort kubernetes resources and helm charts.
E0209 13:08:19.899861       1 gitrepo_sync.go:112] lstat /tmp/data-services/release/internal: no such file or directory
E0209 13:08:19.899888       1 mcmhub_controller.go:626] subscription-hub-reconciler "msg"="failed to process on doMCMHubReconcile" "error"="lstat /tmp/data-services/release/internal: no such file or directory"

Release version:
ACM v2.4.1 (OCP 4.8 (ARO))

Operator snapshot version:

OCP version:
4.8.29 (ARO)

Browser Info:

Steps to reproduce:
1. Configure applications to be deployed via HelmRepo channel
2. Access multicluster_operators_hub_subscription via rsh
3. Check the /tmp folder - observe that subfolders named charts* will be constantly generated

Actual results:
New subfolders charts* will be constantly created in /tmp folder. Even in the moderate environments (6 managed clusters, ~30 apps) /tmp will run out of free space in just few days (~70G, 250k charts* folders in about 4 days)

Expected results:
/tmp usage does not grow

Additional info:
@Roke Jung proposed solution to include CHARTS_DIR environment variable ("/tmp") value which seems to help with the problem.
I think this should be included OOB.