clients(lr): adjust CPU throttling based on benchmark

[adamraine]

Further analysis indicates that 1.2x might be better than 1.5x in PSI. 1.2x is recommended by https://lighthouse-cpu-throttling-calculator.vercel.app/ but 1.5x was just the high end of the reported range.

[connorjclark]

I mentioned elsewhere, but that web app is using a different LHBIDX target than our current moto g4 power 2022 device. it uses 150 as the target score, but if we wrote that app today we would use ~264

Does this modify your suggested throttling rate at all?

[adamraine]

Does this modify your suggested throttling rate at all?

I found the code for the CPU throttling estimator here. Lazily replacing 150 with 264 gives us a multiplier of ~1.05 or ~1.1 if we round it.

That being said, the confidence range of 1.2 contains 1.05 and vice versa so I think it's splitting hairs.

[jmwebservices]

As a performance-focused developer, my PSI mobile performance scores would inconsistently dip into the high 80s due to blocking from third-party scripts. Interestingly, GTM was the only third-party script used. Thankfully, it seems this issue has been resolved, and now my mobile performance scores consistently hit 99-100!