Add WithConnectTimeout to bound and retry session establishment

[r-jais]

What

Adds a WithConnectTimeout(d time.Duration) connection option and a Config.ConnectTimeout field. When set, connector.Connect bounds the whole OpenSession by that deadline and keeps retrying while the failure is transient (e.g. a cold-starting warehouse returning HTTP 503), stopping as soon as the deadline elapses.

Why

A non-serverless Databricks cluster that cold-starts can take minutes to accept a session. Today OpenSession is a single attempt whose retries are capped by the shared RetryMax (default 4), so a slow cold start fails well before the cluster is ready and there is no knob to wait longer. See dbt-labs/dbt-core#14529.

Behavior

• ConnectTimeout <= 0 (default): a single attempt — identical to prior behavior. This is a purely additive mechanism; existing callers are unaffected.
• ConnectTimeout > 0: retry OpenSession with exponential backoff (bounded by RetryWaitMin/RetryWaitMax) until success or the deadline. Retries here are independent of the query-level RetryMax, so query behavior is unchanged.
• Non-retryable failures (bad token, missing warehouse, ...) fail fast — the loop only continues while errors.Is(err, RetryableError).

Tests

TestWithConnectTimeout (option sets the config; defaults to zero). Existing TestNewConnector unaffected.

Part of a cross-repo change

1. this PR (databricks-sql-go) — the mechanism
2. arrow-adbc — Databricks driver databricks.connect_timeout option + default, consuming this
3. dbt-labs/fs — wire the connect_timeout profile field through

The consumer (2) can only bump its go.mod to depend on WithConnectTimeout once this is merged and tagged.