call _get_shunts once not twice

[luke-kiernan]

I ran PProf on assembling the y bus matrix for unrelated reasons and the total time inside _get_shunts was surprisingly high. Fetching the values for both "from" and "to" sides in a single call helps.

Aside: I suspect the cost is primarily due to the type instability in the getters, which will be resolved by my unit fixes.

[Copilot]

Pull request overview

This PR optimizes Y-bus branch entry computation by avoiding repeated shunt getter calls during matrix assembly, reducing time spent in shunt retrieval on profiling.

Changes:

• Replace per-side _get_shunt(br, node) calls with a single _get_shunts(br) call returning both end shunts.
• Cache get_g(br) / get_b(br) results within _get_shunts to avoid repeated getter overhead.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[luke-kiernan]

Pre-existing issue. But yeah, it is strange that we only check Y11 and Y_l, not Y22. @m-bossart is there a reason for this?

[github-actions]

Performance Results

Precompile Time

Main	This Branch	Delta
2.162 s	2.107 s	-2.5%

Execution Time

Test	Main	This Branch	Delta
matpower_ACTIVSg2000_sys-Build PTDF First	1.746 s	1.743 s	-0.2%
matpower_ACTIVSg2000_sys-Build PTDF Second	96.2 ms	71.3 ms	-25.9%
matpower_ACTIVSg2000_sys-Build Ybus First	13.8 ms	116.2 ms	+743.0%
matpower_ACTIVSg2000_sys-Build Ybus Second	13.2 ms	11.0 ms	-16.6%
matpower_ACTIVSg2000_sys-Build LODF First	614.1 ms	527.4 ms	-14.1%
matpower_ACTIVSg2000_sys-Build LODF Second	170.1 ms	156.3 ms	-8.1%
matpower_ACTIVSg2000_sys-Build VirtualMODF First	3.78 s	4.965 s	+31.3%
matpower_ACTIVSg2000_sys-Build VirtualMODF Second	201.7 ms	712.0 ms	+253.1%
matpower_ACTIVSg2000_sys-VirtualMODF Query 10 rows	502.6 ms	496.1 ms	-1.3%
matpower_ACTIVSg2000_sys-Radial network reduction First	470.3 ms	449.9 ms	-4.3%
matpower_ACTIVSg2000_sys-Radial network reduction Second	0.7 ms	0.7 ms	+5.5%
matpower_ACTIVSg2000_sys-Degree two network reduction First	1.852 s	1.74 s	-6.0%
matpower_ACTIVSg2000_sys-Degree two network reduction Second	1.0 ms	0.9 ms	-13.4%
Base_Eastern_Interconnect_515GW-Build Ybus First	3.707 s	3.785 s	+2.1%
Base_Eastern_Interconnect_515GW-Build Ybus Second	3.539 s	3.426 s	-3.2%
Base_Eastern_Interconnect_515GW-Radial network reduction First	35.6 ms	31.7 ms	-10.9%
Base_Eastern_Interconnect_515GW-Radial network reduction Second	38.9 ms	32.5 ms	-16.6%
Base_Eastern_Interconnect_515GW-Degree two network reduction First	378.6 ms	362.2 ms	-4.3%
Base_Eastern_Interconnect_515GW-Degree two network reduction Second	42.2 ms	33.0 ms	-21.7%

[luke-kiernan]

Ok the perf measurements aren't bearing this out--I took Claude's word for it--so I'm going to close this. The root problem is the type instability in the getters, which will go away after the units fixes anyway