deep merge can't nullify `autoscale` settings in target.

[enceladus-beep]

Describe the issue

I'm trying to override the autoscale setting in a specific target to use a single-node cluster.

Configuration

resources:
  jobs: &job-config
    xxx_job: &xxx_job
      name: xxx_job
      tags: &tags
        ver: ver1
      tasks:
        - task_key: tk1
          job_cluster_key: &tk1_cluster_key tk1_cluster
      job_clusters: &job_clusters
        - job_cluster_key: tk1_cluster
          new_cluster: &tk1_cluster
            spark_version: 16.4.x-scala2.13
            spark_conf: &spark_conf
              spark.databricks.delta.stalenessLimit: 20000
              spark.sql.shuffle.partitions: auto
            custom_tags: &cluster_custom_tags
              Team: X-Team
            node_type_id: r5d.large
            autoscale:
              min_workers: 1
              max_workers: 2


bundle:
  name: bundleX

targets:
  t1:
  
  t2: 
    resources:
      jobs:
        xxx_job:
          job_clusters:
            - job_cluster_key: tk1_cluster
              new_cluster:
                node_type_id: r5d.4xlarge
                # nullify autoscale settings on this singleNode cluster
                autoscale: null
                num_workers: 0
                spark_conf:
                  <<: *spark_conf
                  spark.databricks.cluster.profile: singleNode
                  spark.master: "local[*]"
                custom_tags:
                  <<: *cluster_custom_tags
                  Task: apiband-bola-attack
                  ResourceClass: SingleNode

Steps to reproduce the behavior

1. Run a summary of t1 see the default value for autoscale:

$ databricks -p test bundle summary -t t1 -o json | gron | grep new_cluster
json.resources.jobs.xxx_job.job_clusters[0].new_cluster = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale.max_workers = 2;
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale.min_workers = 1;
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags.Team = "X-Team";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.node_type_id = "r5d.large";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.databricks.delta.stalenessLimit"] = "20000";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.sql.shuffle.partitions"] = "auto";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_version = "16.4.x-scala2.13";

2. Run a summary for t2 and see that the autoscale settings are not nullified.

$ databricks -p test bundle summary -t t2 -o json | gron | grep new_cluster
json.resources.jobs.xxx_job.job_clusters[0].new_cluster = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale.max_workers = 2;
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.autoscale.min_workers = 1;
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags.ResourceClass = "SingleNode";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags.Task = "apiband-bola-attack";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.custom_tags.Team = "X-Team";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.node_type_id = "r5d.4xlarge";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.num_workers = 0;
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf = {};
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.databricks.cluster.profile"] = "singleNode";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.databricks.delta.stalenessLimit"] = "20000";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.master"] = "local[*]";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_conf["spark.sql.shuffle.partitions"] = "auto";
json.resources.jobs.xxx_job.job_clusters[0].new_cluster.spark_version = "16.4.x-scala2.13";

Expected Behavior

No autoscale settings in target t2

Actual Behavior

The autoscale setting is deep merged into t2

OS and CLI version

$ databricks --version 
Databricks CLI v1.4.0

Is this a regression?

No.

Debug Logs

Attached to the bug, although not relevant to this issue, it doesn't show anything about deep merge.