fix(python): add timeout parameter to FleetApi.start()

[haosenwang1018]

Problem

FleetApi.start() inherits the default 30s timeout from JsonRpcClient.request(), but session.fleet.start is a blocking RPC that only responds once the fleet completes. Any workload taking more than 30s raises asyncio.TimeoutError (#539).

Fix

Add an explicit timeout keyword parameter to FleetApi.start() (default: 600s) and pass it through to self._client.request().

# Before
await session.rpc.fleet.start(params)  # always 30s timeout

# After
await session.rpc.fleet.start(params)  # 600s default
await session.rpc.fleet.start(params, timeout=1800)  # custom

Fixes #539

[Rasaboun]

It’s a good solution, but there is another problem fleet.start returns started = True after the fleet completes its tasks, not when the fleet is just deployed

[SteveSandersonMS]

Thanks for the proposal here but this is generated code. We can't edit it directly. Changes will be lost. Instead, any changes need to be either in some wrapper API or in the underlying runtime.