feat: v2 swift ipampool

[rbtr]

Reason for Change:

v2 IPAM Pool Monitor introduces idempotent scaling math, migrates to an event-driven instead of polling architecture, and leverages the Pod watcher for IP demand instead of using the "free IPs in the pool" to trigger scaling.

Notably this change improves pool scaling performance from O(n) to O(1).

Testing on a single Linux node DS2v2 with a 100 Pod pause container scale-up yields:

100 pod scaleup test	CNS mode	typical time to full IP availability
	IPAMv1	1:30
	IPAMv2	0:13

Issue Fixed:

Requirements:

• uses conventional commit messages
• includes documentation
• adds unit tests
• relevant PR labels added

Notes: not to be confused with SwiftV2, this is applicable to all dynamic Pod IP Swift scenarios

Design docs for reference: #2013

[nddq]

Left some comments, I'm wondering if we could try and do some stress tests on an actual cluster (or you have done it already 🙂)

[rbtr]

I'm wondering if we could try and do some stress tests on an actual cluster (or you have done it already 🙂)

I don't have numbers, but I validated that it works as expected and reaches target state faster for big swings in Pod scheduling 🙂
The really interesting results will be testing this is combination with @behzad-mir's stateless CNI changes. I expect the time-to-Pod-ready for big Pod scale-ups to be much much lower.

[nddq]

LGTM 🚀

[rbtr]

going to move this one forward, and consider the feedback while running it for real to refine it.