Related to T3350

The first iteration went wrong (not exactly clear what). Nonetheless, the node got
decomissioned in the master and the node got removed and created back after that.

Now everyting is fine.

Related to T3350

Related to T3283

Avoid to override the specific configuration of journal for cloudinit

Related to T3242

Related to T3175

Related to T3159

The end goal is to replicate our current production environment. And making sure writing
to the journal with more than 1 indexer behave properly.

terraform apply summary:
```
worker3_summary =
hostname: worker3
fqdn: worker3.internal.staging.swh.network
network: ip=192.168.130.103/24,gw=192.168.130.1 macaddrs=1A:F8:1A:2C:12:E1
```

Related to T2780

It will allow to manually start several instances
of the swh-search-journal-client

Related to T2935

Related to T2827

Related to T2905

Related to T2905

Related to T2905

Related to T2827

Related to T2827

This inlines the production/common.tf within the production/production.tf

terraform apply output:
```
objstorage0_summary =
hostname: objstorage0
fqdn: objstorage0.internal.staging.swh.network
network: ip=192.168.130.110/24,gw=192.168.130.1 macaddrs=5E:28:EA:7D:50:0D
```

Related to T2682

Superior version will need some upgrade on terraform.tfstate file.
So right now, pin the version to 13.5.

This matches what's currently done for the staging area.

Related to T2888

terraform apply output:
```
clearly-defined_summary =
hostname: clearly-defined
fqdn: clearly-defined.internal.staging.swh.network
network: ip=192.168.130.200/24,gw=192.168.130.1 macaddrs=B6:C6:2E:D3:60:B2
```

Related to T2865

- disk: to be able to retrieve the origin production index
- memory: to lower the pression on the gc during mass import and
  add more fs cache for the searches
- cpu: to have more threads on the ES thread pools (with the default
  configuration). The cpu was the bottleneck during mass import and
  it leads to OOMs eventually

Related to T2817

`terraform apply` result:
```
search0_summary =
hostname: search0
fqdn: search0.internal.staging.swh.network
network: ip=192.168.130.90/24,gw=192.168.130.1 macaddrs=EE:FA:76:55:CF:99
```

Related to T2817

This also modifies the ballooning to 9Gib because the heap is configured to be
half the memory of the machine (16Gib). With a ballooning starting at 1Gib,
elasticsearch refuses to start.

Related to T2817

Output:
```
search-esnode0_summary =
hostname: search-esnode0
fqdn: search-esnode0.internal.staging.swh.network
network: ip=192.168.130.80/24,gw=192.168.130.1 macaddrs=96:74:49:BD:B5:08
```

Related to T2817

Related to T2790

They were decomissioned alongside orsay

Related to T2747

Somehow, it stopped working.
Only seen recently as we needed to create back a new machine.

Related to T2747

They used to be installed in the template. We now inline those directly prior
to the puppet step run.

Related to T2747

Related to T2747