Allow filtering cassandra nodes by datacentre

[jschaul]

Context: https://wearezeta.atlassian.net/browse/BM-14 and related PR https://github.com/zinfra/cailleach/pull/703

• cassandra-util: add a dcAwarePolicy for talking to cassandra
• Spar/brig/galley/gundeck: allow filtering nodes by datacentre

Example: galley/gundeck don't configure this option, spar configures a valid datacentre, brig configures an invalid datacentre:

[galley] I, Using the following cassandra load balancing options ('Policy'):, filter_datacentre=Nothing
[gundeck] I, Using the following cassandra load balancing options ('Policy'):, filter_datacentre=Nothing
[spar] I, Using the following cassandra load balancing options:, filter_datacentre=Just "datacenter1"

[brig] I, Using the following cassandra load balancing options ('Policy'):, filter_datacentre=Just "datacenter2"
[brig] I, logger=cassandra.brig, Known hosts: []
[brig] I, logger=cassandra.brig, New control connection: datacenter1:rack1:127.0.0.1:9042#<socket: 11>
[brig] E, logger=cassandra.brig, No known hosts.
[brig] I, logger=cassandra.brig, Renewing control connection with initial contacts ...

(...)
brig: Failed to read schema version from meta table. Error was: cql-io: no host available
[brig] E, logger=cassandra.brig, No known hosts.
[brig] I, logger=cassandra.brig, Renewing control connection with initial contacts ...

So the behaviour is:

• if not configured, fallback to random policy (i.e. no change from now)
• if configured but wrongly, the service will fail to start up
• if configured correctly, only nodes in the datacentre that is configured will be used as coordinator nodes, so another datacentre going down will have no effect.

Checklist

• The PR Title explains the impact of the change.
• The PR description provides context as to why the change should occur and what the code contributes to that effect. This could also be a link to a JIRA ticket or a Github issue, if there is one.
• changelog.d contains the following bits of information:
  • A file with the changelog entry in one or more suitable sub-sections. The sub-sections are marked by directories inside changelog.d.

[jschaul]

This was dead code not used anywhere.