Refactor Catalyst pass registering

[joeycarter]

Currently we register all Catalyst passes in the Catalyst/Transforms/RegisterAllPasses.cpp source file:

catalyst/mlir/lib/Catalyst/Transforms/RegisterAllPasses.cpp

Lines 24 to 66 in ef4f103

	void catalyst::registerAllCatalystPasses()
	{
	mlir::registerPass(catalyst::createAddExceptionHandlingPass);
	mlir::registerPass(catalyst::createAdjointLoweringPass);
	mlir::registerPass(catalyst::createAnnotateFunctionPass);
	mlir::registerPass(catalyst::createApplyTransformSequencePass);
	mlir::registerPass(catalyst::createArrayListToMemRefPass);
	mlir::registerPass(catalyst::createCatalystBufferizationPass);
	mlir::registerPass(catalyst::createCatalystConversionPass);
	mlir::registerPass(catalyst::createCopyGlobalMemRefPass);
	mlir::registerPass(catalyst::createDetensorizeSCFPass);
	mlir::registerPass(catalyst::createDisableAssertionPass);
	mlir::registerPass(catalyst::createDisentangleCNOTPass);
	mlir::registerPass(catalyst::createDisentangleSWAPPass);
	mlir::registerPass(catalyst::createEmitCatalystPyInterfacePass);
	mlir::registerPass(catalyst::createGEPInboundsPass);
	mlir::registerPass(catalyst::createGradientBufferizationPass);
	mlir::registerPass(catalyst::createGradientConversionPass);
	mlir::registerPass(catalyst::createGradientPreprocessingPass);
	mlir::registerPass(catalyst::createGradientPostprocessingPass);
	mlir::registerPass(catalyst::createGradientLoweringPass);
	mlir::registerPass(catalyst::createHloCustomCallLoweringPass);
	mlir::registerPass(catalyst::createInlineNestedModulePass);
	mlir::registerPass(catalyst::createIonConversionPass);
	mlir::registerPass(catalyst::createMemrefCopyToLinalgCopyPass);
	mlir::registerPass(catalyst::createMemrefToLLVMWithTBAAPass);
	mlir::registerPass(catalyst::createMitigationLoweringPass);
	mlir::registerPass(catalyst::createQnodeToAsyncLoweringPass);
	mlir::registerPass(catalyst::createQuantumBufferizationPass);
	mlir::registerPass(catalyst::createQuantumConversionPass);
	mlir::registerPass(catalyst::createRegisterInactiveCallbackPass);
	mlir::registerPass(catalyst::createRemoveChainedSelfInversePass);
	mlir::registerPass(catalyst::createMergeRotationsPass);
	mlir::registerPass(catalyst::createScatterLoweringPass);
	mlir::registerPass(catalyst::createSplitMultipleTapesPass);
	mlir::registerPass(catalyst::createTestPass);
	mlir::registerPass(catalyst::createIonsDecompositionPass);
	mlir::registerPass(catalyst::createQuantumToIonPass);
	mlir::registerPass(catalyst::createLoopBoundaryOptimizationPass);
	mlir::registerPass(catalyst::createCommuteCliffordTPPRPass);
	mlir::registerPass(catalyst::createCliffordTToPPRPass);
	mlir::registerPass(catalyst::createCommuteCliffordPastPPMPass);
	}

However, as the number of passes and dialects continues to grow, the size of this file may become unwieldy and difficult to maintain. Can we find a better way to register these passes, for instance, next to the definition of the create*Pass functions?

[paul0403]

Just noting that this line in documentation needs to be updated too https://docs.pennylane.ai/projects/catalyst/en/stable/dev/transforms.html#:~:text=Note%20that%20this%20addition%20in%20RegisterAllPasses.cpp%20needs%20to%20happen%20in%20the%20lib/Catalyst/Transforms%20directory%2C%20regardless%20of%20which%20dialect%20your%20pass%20belongs%20to.