[Backport release-1.6] fix(cluster-api): resync capi-operator every minute - #5081

A restarted capi-operator v0.19.0 drops the create events for the
provider Deployments it finds already running
(kubernetes-sigs/cluster-api-operator#1031). A provider whose Ready
condition went False while the operator was down therefore keeps it
until that Deployment changes or the informer resyncs, and the first
resync after a restart lands two periods later because the handler
registers after the informer has listed. With the 10m default that is
about 20 minutes, during which capi-providers-core times out on the
CoreProvider and everything depending on it waits; every further
operator restart starts the clock again.

With a one-minute period the same repro recovered 2m02s after the
operator took its lease, against no recovery within 4m26s with the
default on the same cluster. A resync replays the cache and, with
providers in sync, ends in "No changes detected" and an unchanged
condition, so it writes nothing to the API server; the operator logged
about 15 lines a minute and used 2m CPU and 26Mi of memory.

The upstream chart has no value for the flag, so a package patch that
make update reapplies adds one, and a bats test holds the rendered
argument.

The test is adapted for the release-1.6 runner: hack/test_helper.bash
does not exist here, so the `load test_helper` line is dropped, and the
scratch file falls back to the cozytest.sh workspace. The assertion is
unchanged.

Assisted-by: LLM
(cherry picked from commit c2dab12)
Signed-off-by: Myasnikov Daniil <[email protected]>