Taxonomy service should use an image volume to get DB

[maverbiest]

When reviewing #6916 I found the image volume approach of making the Deacon index available in the raw-reads processing pod very nice:

loculus/.github/workflows/deacon-index.yml

Lines 58 to 78 in 7d48300

	- name: Publish deacon index as an OCI image for the raw-reads-processing image volume
	run: |
	# Push under a stable filename ("deacon.idx") regardless of the dated
	# build filename, so the mounted path in the deployment never changes.
	#
	# This must be a real container image (FROM scratch + COPY), not a bare
	# OCI artifact (e.g. `oras push --artifact-type ...`): Kubernetes'
	# ImageVolume / containerd only support images with a proper image
	# config (rootfs.diff_ids matching the layers). Artifacts with an empty
	# config fail to mount with "mismatched image rootfs and manifest layers".
	# See https://github.com/containerd/containerd/issues/11907
	cp "${INDEX_VERSIONED}" deacon.idx
	cat <<'EOF' > Dockerfile.deacon-index
	FROM scratch
	COPY deacon.idx /deacon.idx
	EOF
	docker buildx build \
	--push \
	--tag "${DEACON_INDEX_IMAGE}:${DATE_TAG}" \
	--tag "${DEACON_INDEX_IMAGE}:latest" \
	-f Dockerfile.deacon-index .

I think we could also use this for making the taxonomy DB available to the taxonomy service. There, we currently run an init container to download the DB from S3:

loculus/kubernetes/loculus/templates/taxonomy-deployment.yaml

Lines 26 to 41 in 7d48300

	initContainers:
	- name: download-taxonomy-db
	image: alpine:3
	command: ["sh", "-c"]
	args:
	- |
	set -e
	DB_URL="https://loculus-public.hel1.your-objectstorage.com/taxonomy/{{ .Values.taxonomyService.dbVersion }}.sqlite.gz"
	echo "Downloading taxonomy DB from $DB_URL"
	wget -q "$DB_URL" -O /data/tmp.sqlite.gz
	gunzip /data/tmp.sqlite.gz
	mv /data/tmp.sqlite /data/taxonomy.sqlite
	echo "Taxonomy DB ready"
	volumeMounts:
	- name: taxonomy-db
	mountPath: /data

From reading a bit, it seems the image volume might have much more preferable caching behabiour at the node level compare to the init container, which makes it so we don't have to download the taxonomy DB (~3GiB) every time we start a new taxonomy service.

Required changes

• Change the action that creates & uploads taxonomy DB to instead create & push a simple docker image with the DB https://github.com/loculus-project/loculus/blob/main/.github/workflows/ncbi-taxonomy-db.yml
• Remove init container from taxonomy-deployment.yaml
• Adapt docs & config in the taxonomy service as appropriate for the new deployment strategy