JIT: allow inlinee profile scale-up

[AndyAyersMS]

We can't always be sure that callee entry counts will be greater than or equal
to call site counts. For example a callee could be tiered up before a caller,
and so have relatively smaller counts.

So, when computing inline scale, allow callee counts to be scaled up as well as
scaled down.

[AndyAyersMS]

cc @dotnet/jit-contrib

Local testing with special Tiered PGO SPMI collection shows a few diffs:

Total bytes of delta: -170 (-1.72% of base)
    diff is an improvement.


Top file regressions (bytes):
          11 : 77033.dasm (1.45% of base)
           6 : 77059.dasm (0.79% of base)

Top file improvements (bytes):
        -160 : 77248.dasm (-14.32% of base)
         -12 : 99351.dasm (-0.67% of base)
         -10 : 76990.dasm (-1.87% of base)
          -3 : 77035.dasm (-0.35% of base)
          -1 : 99696.dasm (-0.05% of base)
          -1 : 77061.dasm (-0.11% of base)

8 total files with Code Size differences (6 improved, 2 regressed), 1 unchanged.

Top method regressions (bytes):
          11 ( 1.45% of base) : 77033.dasm - EMFloat:MultiplyInternalFPF(byref,byref,byref)
           6 ( 0.79% of base) : 77059.dasm - EMFloatClass:MultiplyInternalFPF(InternalFPF,InternalFPF,InternalFPF)

Top method improvements (bytes):
        -160 (-14.32% of base) : 77248.dasm - LUDecomp:DoLUIteration(System.Double[][],System.Double[],System.Double[][][],System.Double[][],int):long
         -12 (-0.67% of base) : 99351.dasm - DynamicClass:ReadJsonBenchmarks.TestObjectFromXml(System.Runtime.Serialization.XmlReaderDelegator,System.Runtime.Serialization.XmlObjectSerializerReadContext,System.Xml.XmlDictionaryString[],System.Xml.XmlDictionaryString[]):System.Object
         -10 (-1.87% of base) : 76990.dasm - StringSort:LoadStringArray(System.String[][],int,int)
          -3 (-0.35% of base) : 77035.dasm - EMFloat:DivideInternalFPF(byref,byref,byref)
          -1 (-0.05% of base) : 99696.dasm - DynamicClass:ReadJsonBenchmarks_TestObjectFromJson(System.Runtime.Serialization.XmlReaderDelegator,System.Runtime.Serialization.Json.XmlObjectSerializerReadContextComplexJson,System.Xml.XmlDictionaryString,System.Xml.XmlDictionaryString[]):System.Object
          -1 (-0.11% of base) : 77061.dasm - EMFloatClass:DivideInternalFPF(InternalFPF,InternalFPF,InternalFPF)

Top method regressions (percentages):
          11 ( 1.45% of base) : 77033.dasm - EMFloat:MultiplyInternalFPF(byref,byref,byref)
           6 ( 0.79% of base) : 77059.dasm - EMFloatClass:MultiplyInternalFPF(InternalFPF,InternalFPF,InternalFPF)

Top method improvements (percentages):
        -160 (-14.32% of base) : 77248.dasm - LUDecomp:DoLUIteration(System.Double[][],System.Double[],System.Double[][][],System.Double[][],int):long
         -10 (-1.87% of base) : 76990.dasm - StringSort:LoadStringArray(System.String[][],int,int)
         -12 (-0.67% of base) : 99351.dasm - DynamicClass:ReadJsonBenchmarks.TestObjectFromXml(System.Runtime.Serialization.XmlReaderDelegator,System.Runtime.Serialization.XmlObjectSerializerReadContext,System.Xml.XmlDictionaryString[],System.Xml.XmlDictionaryString[]):System.Object
          -3 (-0.35% of base) : 77035.dasm - EMFloat:DivideInternalFPF(byref,byref,byref)
          -1 (-0.11% of base) : 77061.dasm - EMFloatClass:DivideInternalFPF(InternalFPF,InternalFPF,InternalFPF)
          -1 (-0.05% of base) : 99696.dasm - DynamicClass:ReadJsonBenchmarks_TestObjectFromJson(System.Runtime.Serialization.XmlReaderDelegator,System.Runtime.Serialization.Json.XmlObjectSerializerReadContextComplexJson,System.Xml.XmlDictionaryString,System.Xml.XmlDictionaryString[]):System.Object

8 total methods with Code Size differences (6 improved, 2 regressed), 1 unchanged.

LU decomp shows exactly this case...

Invoking compiler for the inlinee method LUDecomp:lusolve(System.Double[][],int,System.Double[]):int :
...
Computing inlinee profile scale:
   call site count 861.000000 callee entry count 115.000000 scale 7.486957

[briansull]

Looks Fine to me