need to tune cluster sylvactl timeout for libvirt-metal case

We have code to compute the sylvactl per-unit timeout for cluster unit, based on the number of nodes and on heuristic timing hints:

# pass the values for node_creation and node_drain timeout in minutes
sylvactl_timing_hints:
  node_create_timeout: >-
    {{ ternary 30 15 (eq .Values.cluster.capi_providers.infra_provider "capm3") | include "preserve-type" }}
  node_drain_timeout: >-
    {{ ternary 30 15 (eq .Values.cluster.capi_providers.infra_provider "capm3") | include "preserve-type" }}

In libvirt-metal CI, these timings are very pessimistic because libvirt-metal boots much faster that real baremetal.

This results, in CI runs that fail on this unit, in having a uselessly very long time for job timeout, which leads to having to wait for result feedback and also consumes more CI resources.

We need to reduce these timeouts for these runs.

I think we should do it at the sylva-units level to conditionally use lower timeouts if libvirt-metal unit is enabled.