[Manager] Optimize watch instance deployment states implementation & Add enable_port_offset_store arg

[s5u13b]

1. Previously, when the WATCH_DEPLOYMENT_INTERVAL is not large enough, manager might scale down the initializing instance wrongly. In this pr, when instance is initializing (PENDING_CREATION), the manager will sleep WATCH_DEPLOYMENT_INTERVAL_PENDING_INSTANCE (longer than WATCH_DEPLOYMENT_INTERVAL) to avoid scaling down the the initializing instance wrongly.
2. Add enable_port_offset_store arg to avoid bench test waiting timeout due to recovering old port offset from ray cluster kv store.

[ZeldaHuang]

LGTM

[github-actions]

prefill	p25	p50	p75	p95	p99	mean
latency(ms)	16835.55	77032.66	142973.14	166469.29	180755.01	80882.11

decode	p25	p50	p75	p95	p99	mean
latency(ms)	50.53	53.66	61.03	93.13	203.74	62.94

[github-actions]

prefill	p25	p50	p75	p95	p99	mean
latency(ms)	13087.11	76980.83	119137.23	196773.45	198203.94	78585.50

decode	p25	p50	p75	p95	p99	mean
latency(ms)	49.90	54.70	65.74	108.06	268.95	70.14

[github-actions]

migration_size	8.00 MB	16.00 MB	24.00 MB	32.00 MB	40.00 MB	48.00 MB	56.00 MB	64.00 MB	72.00 MB	80.00 MB	88.00 MB	96.00 MB	104.00 MB	112.00 MB	120.00 MB	128.00 MB	136.00 MB	144.00 MB	152.00 MB	160.00 MB	168.00 MB	176.00 MB	184.00 MB	192.00 MB	200.00 MB	208.00 MB	216.00 MB	224.00 MB	232.00 MB	240.00 MB	256.00 MB	280.00 MB	304.00 MB	408.00 MB	464.00 MB
rayrpc_speed(GB/s)	1.04	1.54	1.78	1.95	2.06	2.12	2.18	2.21	2.25	2.27	2.41	2.40	2.42	2.43	2.42	2.51	2.55	2.58	2.56	2.44	2.56	2.64	2.66	2.54	2.58	2.87	2.78	2.86	2.74	2.97	2.78	2.57	3.07	3.36	3.14

migration_size	8.00 MB	16.00 MB	24.00 MB	32.00 MB	40.00 MB	48.00 MB	56.00 MB	64.00 MB	72.00 MB	80.00 MB	88.00 MB	96.00 MB	104.00 MB	112.00 MB	120.00 MB	128.00 MB	136.00 MB	144.00 MB	152.00 MB	160.00 MB	168.00 MB	176.00 MB	184.00 MB	192.00 MB	200.00 MB	208.00 MB	232.00 MB	264.00 MB
gloo_speed(GB/s)	1.01	1.66	2.10	2.33	2.60	2.89	2.98	2.84	2.94	3.10	3.16	3.20	3.20	3.78	3.25	2.76	2.71	2.63	2.73	3.32	1.97	3.13	2.17	2.07	2.51	3.24	1.93	2.83