Convert no longer necessary lambda expressions to method groups

[Joehuu]

Noticed in passing. Mentioned Issues are closed, but I haven't tested performance.

[bdach]

but I haven't tested performance

Please benchmark. KeyBindingContainer is a hot path and must keep allocs low. See #5796.

99% of the work in this PR is verifying that the dotnet issue mentioned in comments is fixed, not applying the obvious change. I'm kind of at the point where I'm no longer willing to suck it up and just test it myself.

[Joehuu]

Just ran what was in dotnet/runtime#33747, and it seems fixed:

Method	Mean	Error	StdDev	Allocated
ByMethodGroup	452.7 ns	4.44 ns	4.16 ns	-
ByDelegate	457.6 ns	7.48 ns	7.35 ns	-

[bdach]

Please benchmark directly via our direct usage site, not via that reproducer. I'm not trying to be unreasonable, compilers are weird and sometimes they fail to optimise if an optimisation site is too indirect to be recognisable for the optimisation.

[Joehuu]

During .NET 6 times to compare (checked out 2c9e7a6, and reverted that commit):

Method	Mean	Error	StdDev	Gen0	Allocated
ByDelegate	1.763 ms	0.0058 ms	0.0051 ms	3007.8125	6 MB
ByMethodGroup	1.756 ms	0.0053 ms	0.0044 ms	3039.0625	6.06 MB

master:

Method	Mean	Error	StdDev	Gen0	Gen1	Allocated
ByDelegate	1.397 ms	0.0273 ms	0.0268 ms	1009.7656	13.6719	6.05 MB
ByMethodGroup	1.442 ms	0.0287 ms	0.0402 ms	1009.7656	13.6719	6.05 MB

Manually renamed to ByMethodGroup on the table with the method group change for display purposes.

A janky benchmark with unrealistic input order, doesn't hit the line if ctrl is first then f1 (adding [Test] attribute and debugging test seems to breakpoint at the line I'm benchmarking):

using BenchmarkDotNet.Attributes;
using osu.Framework.Input;
using osu.Framework.Input.Events;
using osu.Framework.Input.States;
using osuTK.Input;

namespace osu.Framework.Benchmarks
{
    [MemoryDiagnoser]
    public partial class MethodGroupBenchmarks
    {
        [Benchmark]
        public void ByDelegate()
        {
            // use FrameworkActionContainer as it has DefaultKeyBindings
            var keyBindingContainer = new TestFrameworkActionContainer();

            for (int i = 0; i < 1000; i++)
            {
                keyBindingContainer.TriggerEvent(new KeyDownEvent(new InputState(), Key.F1));
                keyBindingContainer.TriggerEvent(new KeyDownEvent(new InputState(), Key.ControlLeft));
                keyBindingContainer.TriggerEvent(new KeyUpEvent(new InputState(), Key.F1));
                keyBindingContainer.TriggerEvent(new KeyUpEvent(new InputState(), Key.ControlLeft));
            }
        }

        private partial class TestFrameworkActionContainer : FrameworkActionContainer
        {
            public TestFrameworkActionContainer()
            {
                ReloadMappings();
            }
        }
    }
}

This is my first time benchmarking something, so sorry if this isn't the proper way.