recovery-point-manager role

The built-in roles.recovery-point-manager role periodically creates
recovery points (see box.backup.recovery_point.create()) across a
cluster from the declarative configuration. Enable it via roles and
configure it under roles_cfg.

Role-level options (defaults inherited by every manager):

• backend (string) -- the default backend type. Resolved to the
  module roles.recovery-point-manager.backends..
• backends_cfg (map) -- per-backend-type config, keyed by the backend
  type name (for example, net-replicaset). The value is that
  backend's own configuration.
• create.by.interval (number) -- seconds between attempts
  (default 600).
• create.timeout (number) -- per-attempt timeout in seconds
  (default 300).
• managers (map) -- the managers to run, keyed by a user-chosen name.
  Omitting managers runs a single manager named default; an
  explicit empty map runs none.

Per-manager options (under managers.<name>), each overriding the
role-level default:

• backend (string) -- the backend type for this manager.
• backend_cfg (table) -- the configuration of this manager's backend
  type (un-keyed, unlike the role-level backends_cfg). It deep-merges
  onto the role-level backends_cfg[<backend>], the manager winning.
• create.by.interval (number) -- seconds between attempts.
• create.timeout (number) -- per-attempt timeout in seconds.

The built-in net-replicaset backend creates a recovery point on a
remote replicaset's leader over net.replicaset. Its backend_cfg:

• target (string, required) -- the name of the target replicaset in
  the cluster configuration.
• login (string) -- the user the backend connects as. When omitted,
  the identity is taken from iproto.advertise.sharding, then
  iproto.advertise.peer.
• password (string) -- the password for login. When omitted, it is
  resolved from the credentials of that user. Cannot be set without
  login.
• params (table, Enterprise Edition) -- the connection transport:
  transport, ssl_key_file, ssl_cert_file, ssl_ca_file, ssl_ciphers,
  ssl_password, ssl_password_file.

The login user needs only the minimum privilege to create recovery
points on the target: execute on the box.backup.recovery_point.create
lua_call. A dedicated user with nothing else is the recommended setup:

credentials:
  users:
    backup_user:
      password: secret_backup
      privileges:
        - permissions: [execute]
          lua_call: ['box.backup.recovery_point.create']

Example:

roles: [roles.recovery-point-manager]
roles_cfg:
  roles.recovery-point-manager:
    backend: net-replicaset
    create:
      by:
        interval: 300
    backends_cfg:
      net-replicaset:
        login: backup_user
    managers:
      rs_001:
        backend_cfg:
          target: replicaset-001

Requested by @Serpentian in https://github.com/tarantool/tarantool/commit/9398bba1d7284cccba00a7518feee16d0073d57a.

Start from the recovery-point-manager role and the box.backup.recovery_point.create entry point described in the issue. Document the role-level and per-manager options, net-replicaset backend settings, required privilege, and YAML example; done means the configuration behavior and example are covered clearly in the relevant documentation.