Constprop through chlo.lgamma

[vimarsh6739]

[vimarsh6739]

Turns out, just calling the lowering to stablehlo isn't enough, we also need to const prop log and log_plus_one.

    %cst = stablehlo.constant dense<0x7F800000> : tensor<f32>
    %cst_0 = stablehlo.constant dense<0.918938517> : tensor<f32>
    %cst_1 = stablehlo.constant dense<2.01490307> : tensor<f32>
    %cst_2 = stablehlo.constant dense<7.500000e+00> : tensor<f32>
    %cst_3 = stablehlo.constant dense<1.000000e+00> : tensor<f32>
    %cst_4 = stablehlo.constant dense<5.000000e-01> : tensor<f32>
    %cst_5 = stablehlo.constant dense<0.000000e+00> : tensor<f32>
    %cst_6 = stablehlo.constant dense<263.38324> : tensor<f32>
    %0 = stablehlo.log_plus_one %cst_5 : tensor<f32>
    %1 = stablehlo.add %cst_1, %0 : tensor<f32>
    %2 = stablehlo.divide %cst_2, %1 : tensor<f32>
    %3 = stablehlo.subtract %cst_4, %2 : tensor<f32>
    %4 = stablehlo.multiply %3, %1 : tensor<f32>
    %5 = stablehlo.log %cst_6 : tensor<f32>
    %6 = stablehlo.add %cst_0, %4 : tensor<f32>
    %7 = stablehlo.add %6, %5 : tensor<f32>
    %8 = stablehlo.abs %cst_3 : tensor<f32>
    %9 = stablehlo.compare  EQ, %8, %cst : (tensor<f32>, tensor<f32>) -> tensor<i1>
    %10 = stablehlo.select %9, %cst, %7 : tensor<i1>, tensor<f32>
    return %10 : tensor<f32>