Fix fast patching potentially corrupting EH catch type in memory

[wondercrash]

When fast patching method bodies, the EH catch type token is rewritten directly into the original raw body's section data

AsmResolver/src/AsmResolver.DotNet/Code/Cil/CilMethodBodySerializer.cs

Lines 188 to 199 in 08fa5cd

	var raw = (CilRawFatMethodBody) sourceBody.OriginalRawBody;
	foreach (var section in raw.ExtraSections)
	{
	byte[] sectionData = section.Data;
	if (section.IsEHTable)
	{
	try
	{
	#if NETSTANDARD2_1_OR_GREATER || NETCOREAPP2_1_OR_GREATER
	FastCilReassembler.PatchExceptionHandlerSection(sectionData, tokenRewriter, section.IsFat);
	#else

Since sectionData is ultimately owned by the sourceBody.OriginalRawBody, patching the catch type token in place mutates the original body itself in memory. So for example if you write a module twice, the first module will be fine while the second module may have the wrong token for the catch type.

repro case:

void test()
{
    try
    {
        throw new Exception();
    }
    catch (Exception)
    {
        return;
    }
}

var module = ModuleDefinition.FromModule(typeof(Program).Module);
module.Write(new MemoryStream());

var method = module.TopLevelTypes.SelectMany(t => t.Methods).First(m => m.Name.Contains("test"));

// should print System.Exception, but it doesn't
Console.WriteLine(method.CilMethodBody!.ExceptionHandlers[0].ExceptionType); 

To fix it, I just Clone the sectionData before patching it.

[Washi1337]

Thanks for the catch and the fix. Could you add your repro as a unit test?