[copybara-service]

feat(interactions): add automatic_continuation to interactions.create in the Python SDK

Foreground interactions.create() calls are capped per request by the server (32k output tokens unary, 64k streaming) and stop with status='requires_continuation' plus a continuation_token. GeminiNextGenInteractions.create and AsyncGeminiNextGenInteractions.create now accept automatic_continuation (default True) and chain those continuation hops client-side so a plain create() returns the full decode.

Request chaining:

• Each hop after the first sends the continuation_token from the previous hop. With store=True (default) it sets previous_interaction_id to the hop that issued the token and omits input; with store=False it resends the original input.
• Unary foreground model calls default generation_config.max_output_tokens to 65536 when unset (streaming is left to the server's 64k default). Explicit values are never overridden.
• Stops on completed, incomplete (token preserved for manual resumption), requires_action, failed, cancelled, a missing token, or a repeated token. Skipped for agent / agent_config requests (server-managed loops) and background=True; background=True with automatic_continuation=False raises ValueError.

Unary (stream=False) merging:

• Hops are merged into one Interaction: steps concatenated in hop order (leading user_input steps dropped on later hops), usage counts and per-modality breakdowns summed across hops with usage.initial_input_tokens taken from hop 1, id / status / continuation_token / updated from the final hop, and previous_interaction_id / created / input from hop 1.
• continuation_ids lists every hop id in order (the last equals id), de-duplicated with any server-provided ids. output_text skips intervening thought steps across hop boundaries.

Streaming (stream=True):

• Events from all hops are yielded verbatim on the returned Stream / AsyncStream. Consumers should treat interaction.completed with status='requires_continuation' as intermediate.
• Mid-hop transport drops or premature EOF are resumed from the latest interaction.status_update checkpoint token when no content has been yielded since that checkpoint, with at most 3 no-progress reconnects. SSE error events are terminal. HTTP status retries stay with http_options.retry_options.

Timeouts and errors:

• A numeric per-call timeout bounds the whole create() call across hops: each hop is sent the remaining budget and no further hop starts once it expires. http_options.timeout, httpx.Timeout and http_options.retry_options continue to apply per hop.
• A failure (or timeout) on a hop after earlier hops already produced output raises errors.ContinuationInterruptedError (underlying error as __cause__) carrying partial_interaction (unary: the hops merged so far, status requires_continuation), continuation_token, interaction_id (to send as previous_interaction_id) and continuation_ids, so billed work is not lost and the call can be resumed manually. Failures on the first hop propagate unchanged.

Opt-out:

• automatic_continuation=False returns the single server response unchanged (no budget injection, no merging). A one-time UserWarning is emitted when such a response stops at requires_continuation and the caller did not supply a continuation_token themselves.

Tests: tests/interactions/test_automatic_continuation.py (unit and edge cases, sync and async) and tests/interactions/test_continuation_conformance.py (cross-language conformance suite).