Define per-backend connection schemas (Spark, Postgres, S3, Kafka…)

[rafafrdz]

Motivation

Depends on #1 (top-level `connections:` section).

Once `connections:` exists, each connection `type` needs a well-defined schema so validators, UIs, and runtimes can agree on required/optional fields.

Proposal

Define a reference schema per backend, starting with the ones we already support or plan to support short-term:

`spark`

field	type	required	notes
url	string	✓	`sc://host:port/`
app_name	string		default: "teckel"
connect_timeout_seconds	int		default: 30
request_timeout_seconds	int		default: 300
auth_token	string		bearer token if the cluster requires it

`postgres`

field	type	required
url	string	✓
user	string
password	string
schema	string

`s3`

field	type	required
bucket	string	✓
region	string
access_key_id	string
secret_access_key	string
endpoint	string

`kafka`

field	type	required
bootstrap_servers	string	✓
security_protocol	enum
sasl_mechanism	enum
sasl_username	string
sasl_password	string

Scope

• Document each schema in the spec.
• Encode as JSON Schema so validators can check `connections.*` entries at pipeline-load time.
• Mark fields that support `${VAR}` interpolation vs. literal-only.

Out of scope

• Runtime implementation (lives in teckel-api).
• UI editor (lives in teckel-ui).