* Add compilation support for P4Smith for DPDK, eBPF, V1model, and p4test.

Change $TEST_DIR/out.p4 to $TEST_DIR/out_$i.p4 to generate different programs with different seeds (to inspect the behavior of the generation and compilation process in a finer/smaller granularity)

Switch the order of --arch and --target

Revert the version of backends/p4tools/modules/smith/scripts/compilation-test.sh back to the commit d34f60e, and switch the order of core and generic in backends/p4tools/modules/smith/targets/generic/test/P4Tests.cmake

Modify the argument settings for multiple targets and architectures

Attempt to make sure that only primitive-type parameters are used when generating the parameter list for the BMv2 portion of the fuzzer

Fix the previous erroneous treatment of "bmv2" from archName to deviceName

Add TODOs to remind myself to revert `out_$i.p4` back to `out.p4` to
avoid dumping too many generated files after the debugging process

1. Instead of adding target-specific check in the target-independent core, preliminarily override the virtual method in the BMv2 backend (still hasn't worked yet)

2. Get rid of  and expand the function signature of

Refactor the implemented inheritance pattern

Clang-format some source file

Fix a linter issue

Implement a mechanism to maintain a list of "known" bugs and not fail the test if those bugs occur

Modify the output-printing mechanism

Add another element to `KNOWN_BUGS`

Add an echo statement to display the encounted bug using the `is_known_bug` function

Modify echo statements

Preliminarily add code to generate an approriate return statement for `pna_pre_output_metadata_t`

Fix a linter issue

Add a TODO to fix the casting bug

Make use of `new IR::Type_Name`

Hard-code to attempt to resolve the casting issue

The current type casting is still erroneous

Add commented-out code that doesn't work and add TODOs

Fix linter issues

Resolve the infinite recursing issue

Improve the implementation to be modular (but still doesn't work)

Fix some auxillary issues

Refine type resolving

Refine recursive type resolving

Fix the "Arch '<null>' not supported by P4Runtime serializer" error

Specify the compiler binary with extentions in the corresponding P4Tests.cmake file instead of the bash script

Simplify the constructor Bmv2V1ModelDeclarationGenerator

Remove 128 from the accepted bitwidths in the Expression Generator for the dpdk taget

Activate `byte_align_headers` to `true` to support bit alignments

Add a known bug: `terminate called after throwing an instance of 'Util::CompilerBug'`

Modify the previously documented known bug to a fine-grained version

Make further attempts to resolve type resolving

Resolve a typo made previously

Further refine type resolving and remove redundant code

Implement explicit casting for `SecurityAssocId_t` for both assignment and return statements

Revert the type signature of `genStructListExpr` from `IR::Expression` back to `IR::ListExpression`

Revert the type signature of `getTypeByName` from `IR::Type` back to `IR::Type_Declaration`

Revert the type signature of `genStructListExpr` from `IR::Expression` back to `IR::ListExpression`

Preliminarily modify the previous type resolving/casting mechanism

Remove the previous added, redundant `genBitType` with different arguments than the default `genBitType`

Add print-outs to address the tuple-wrapping issue

Resolve the typing issue of `SecurityAssocId_t` in a more generic way

Make the bug documented in `KNOWN_BUGS` more generic

Experimentally migrate the `Bmv2V1ModelDeclarationGenerator` class and its constructor and (overridden) member function to declarations.cpp and declarations.h in the `bmv2` target

Experimentally migrate the `Bmv2V1ModelDeclarationGenerator` class and its constructor and (overridden) member function to declarations.cpp and declarations.h in the `bmv2` target

Fix linter issues from the previous commit(s)

Fix the issue caused by erroneously formatting backends/p4tools/modules/smith/targets/bmv2/CMakeLists.txt

Set `byte_align_headers` true for the `nic` target

Prohibit `128` from being a valid bit size for the `nic` target

Document a known bug in the script

Fix the previous linter issue(s)

Support to generate well-typed parameter lists for the `nic` target

Remove the hard-coded rhs-casting(s) of assignment generations

Add a TODO (to make the bitwidth a generic parameter)

1. Preliminarily make the bitwidth a generic parameter (by implementing a template function)

2. Document a known bug

Revert backends/p4tools/modules/smith/scripts/compilation-test.sh back to its original form since it is cluelessly `clang-format`ed

Reactivate the executable permission of backends/p4tools/modules/smith/scripts/compilation-test.sh

Preliminarily implement `pna_main_parser_input_metadata_t`'s field generations (although `enum` is not supported right now)

Rename the previous `Bmv2V1ModelDeclarationGenerator` class to `Bmv2ModelDeclarationGenerator` and apply it (and its members) to the `psa` arch

1. Add a struct member in the `Constraints` struct to limit the maximum bit width (default value: 128)

2. Set it to 64 for the pna arch, dpdk target

Fixups.

Signed-off-by: zzmic <100326374+zzmic@users.noreply.github.com>
Signed-off-by: fruffy <fruffy@nyu.edu>

* Clean up initialization of probabilities.

Signed-off-by: fruffy <fruffy@nyu.edu>

* Add support for parallel test runs, simplify bit literal generation.

Signed-off-by: fruffy <fruffy@nyu.edu>

---------

Signed-off-by: zzmic <100326374+zzmic@users.noreply.github.com>
Signed-off-by: fruffy <fruffy@nyu.edu>
Co-authored-by: fruffy <fruffy@nyu.edu>