Fixing Compiler bug

[MathewZach123]

This PR fixes #5549 caused due to DAG-cached stale references in FunctionsInliner by modifying DoRemoveActionParameters to handle inout parameters separately. A new ExtractArrayIndices Transform extracts side-effecting array indices into temporary variables before generating copy-in/copy-out, so the function call is evaluated once.

Updated IR for Test 1 (testdata/p4_16_samples/inline-function-to-array-index.p4)

   @noWarn("unused") @name(".NoAction") action NoAction_0() {
   }
   @name("hdr") h_t hdr_1;
   bit<3> tmp_idx;
   @name("a") action a_0() {
       tmp_idx = max(3w0, 3w0);
       hdr_1 = h.hdr_arr[tmp_idx];
       h.hdr_arr[tmp_idx] = hdr_1;

   }

Updated IR for Test 2 (testdata/p4_16_samples/inline-function-to-array-index-chained.p4)

    @noWarn("unused") @name(".NoAction") action NoAction_0() {
    }
    @name("val") bit<8> val_1;
    bit<3> tmp_idx;
    bit<3> tmp_idx_0;
    @name("a") action a_0() {
        tmp_idx = max(3w0, 3w0);
        tmp_idx_0 = max(3w0, 3w0);
        val_1 = h.mid_arr[tmp_idx].inner_arr[tmp_idx_0].f2;
        val_1 = 8w1;
        h.mid_arr[tmp_idx].inner_arr[tmp_idx_0].f2 = val_1;
    }
 ``

[ChrisDodd]

Actaully, looking at this a bit more, while this fixes this case, I'm not sure this is the correct fix overall -- it is fixing the inlining of the duplicated method call by inlining it twice, but the duplication of the method call might be problematic if the method itself contains side effects.

Not sure it is the right fix -- still reviewing

[ChrisDodd]

See the comment I added to #5549

[MathewZach123]

Thanks @ChrisDodd

[kfcripps]

@MathewZach123 You should be able to remove this line if you rebase: https://github.com/p4lang/p4c/blob/main/backends/p4tools/modules/smith/scripts/compilation-test.sh#L15