Clean up kernel creation code in `codegen.driver`. Fix and dry-test SYCL codegen.

!449 (merged) accidentally broke SYCL code generation since SYCL kernels do not receive a launch config, but their kernel object expected one since it was being modelled as a GPU kernel.

This MR fixes this, alongside a few other improvements:

• Target.SYCL is no longer a GPU target
• SyclPlatform no longer inherits from platforms.GenericGpu, but directly from platforms.Platform.
• Emit SYCL kernels using codegen.Kernel instead of codegen.GpuKernel
• Combine the various kernel creation utility functions in codegen.driver into a KernelFactory class for cleaner code
• Move creation of Lambdas from Lambda.from_expression to KernelFactory.create_lambda
• Add dry-dock tests for SYCL code generation that only test if the codegen driver runs successfully

Relates to #126

src/pystencils/codegen/driver.py

@@ -436,23 +436,11 @@ class DefaultKernelCreationDriver:
                     f"No platform is currently available for CPU target {self._target}"
                 )
 
-        elif Target._GPU in self._target:
+        elif self._target.is_gpu():
             gpu_opts = self._cfg.gpu
             omit_range_check: bool = gpu_opts.get_option("omit_range_check")
 
             match self._target:
-                case Target.SYCL:
-                    from ..backend.platforms import SyclPlatform
-
-                    auto_block_size: bool = self._cfg.sycl.get_option(
-                        "automatic_block_size"
-                    )
-
-                    return SyclPlatform(
-                        self._ctx,
-                        omit_range_check=omit_range_check,
-                        automatic_block_size=auto_block_size,
-                    )
                 case Target.CUDA:
                     from ..backend.platforms import CudaPlatform
 
@@ -467,6 +455,17 @@ class DefaultKernelCreationDriver:
                         omit_range_check=omit_range_check,
                         thread_mapping=thread_mapping,
                     )
+        elif self._target == Target.SYCL:
+            from ..backend.platforms import SyclPlatform
+
+            auto_block_size: bool = self._cfg.sycl.get_option("automatic_block_size")
+            omit_range_check = self._cfg.gpu.get_option("omit_range_check")
+
+            return SyclPlatform(
+                self._ctx,
+                omit_range_check=omit_range_check,
+                automatic_block_size=auto_block_size,
+            )
 
         raise NotImplementedError(
             f"Code generation for target {self._target} not implemented"