feat(compression): allow configuring the Kopia compression policy

[gabriele-wolfox]

What

Kopia repositories were created without a compression policy, so base backup
data was stored uncompressed. This lets users configure Kopia compression for
base backups, both globally and per cluster.

Configuration

Compression can be set at two levels, following Kopia's policy inheritance
(a per-cluster policy overrides the global one for that cluster's source):

• Global (repository-wide) on the Server, under spec.tier1.compression
  and spec.tier2.compression. Applied to Kopia's global policy when the
  server starts.
• Per cluster on the PluginConfiguration, under spec.tier1.compression
  and spec.tier2.compression. Applied to the cluster's own source.

Each policy exposes:

• algorithm: any Kopia compressor (e.g. zstd, s2-default, gzip) or
  none to disable compression.
• minSize / maxSize (optional, bytes): restrict compression to files within
  a size range.

Both tier1 and tier2 are covered. Compression takes effect on the next backup;
existing backups are not recompressed.

Example:

apiVersion: klio.cnpg.io/v1alpha1
kind: PluginConfiguration
spec:
  clusterName: cluster-example
  tier1:
    compression:
      algorithm: zstd
      minSize: 4096
  tier2:
    enableBackup: true
    compression:
      algorithm: zstd-better-compression
      minSize: 4096

Out of scope

Kopia's only-compress / never-compress extension lists are not exposed:
they select files by extension, which does not fit a PostgreSQL data directory
(relation files are numeric), and WAL files are compressed separately.

Testing

A new e2e (compression_test.go) sets a global policy on the Server and a
different per-cluster policy on the PluginConfiguration, runs a backup, and
asserts via kopia policy show that both the --global and user@host
policies carry the expected algorithm and minSize on tier1 and tier2. Docs,
CRDs, Helm chart, and generated API reference are updated.

Closes #126

Assisted-by: Claude