Improve fleet integration management

[Kushmaro]

Situation 1: Mandatory Definition of All Integration Inputs

Customer is encountering an "issue" where they must explicitly define all integration inputs within their Fleet Terraform configurations, even for those that are not enabled. This is necessary to prevent Terraform from showing diffs between plans. For example, in the AWS integration, which offers 50 different inputs, they only enable 4. Despite this, they need to define every input, importing their default values, to avoid plan discrepancies.

Here's an example of the issue with a disabled input still needing to be defined:

resource "elasticstack_fleet_integration_policy" "server-system" {
  name              = "system"
  namespace         = "default"
  description       = "example"
  agent_policy_id   = elasticstack_fleet_agent_policy.server_policy.id
  integration_name  = elasticstack_fleet_integration.this["system"].name
  integration_version = elasticstack_fleet_integration.this["system"].version

  input {
    enabled       = true
    input_id      = "system-logfile"
    streams_json  = jsonencode(jsondecode(file("integration_configs/server/system/system-logfile.json")))
  }
  ...
  input {
    enabled       = false
    input_id      = "system-httpjson"
    streams_json  = jsonencode(jsondecode(file("integration_configs/server/system/system-httpjson.json")))
  }
}

If customer remove the last input and its associated streams, the plan continuously shows a diff suggesting the removal of the undefined input. This requires customer to "backport" the default streams for all inputs, even disabled ones, and to update them with each integration version change.

[psypuff]

This seems to also affect multiple streams within a single input (e.g. the Prometheus integration).
Would love to omit the disabled-by-default streams from the streams_json argument. Not sure how easy would that be given this is a json string and probably not part of the schema. Maybe worth having streams as blocks just for the enabled argument and the rest of their config could be kept as json.