Improve error message when defining custom type for variables

[wangyems]

resolve #4122

The unhelpful message stuff is a bit different from the above link's description. ML.NET throws unhelpful message not because the customer uses the wrong type(different from the type defined in onnx), but defining the variable using the same type as OnnxSequenceType. However, the correct type should be IEnumerable.

This change adds more information to existing exception message and adds a special error message for errors like this when customer carelessly defines a container variable without using container type

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@b74cacc). Click here to learn what that means.
The diff coverage is 0.00%.

@@            Coverage Diff            @@
##             master    #5114   +/-   ##
=========================================
  Coverage          ?   75.56%           
=========================================
  Files             ?      995           
  Lines             ?   179296           
  Branches          ?    19295           
=========================================
  Hits              ?   135488           
  Misses            ?    38544           
  Partials          ?     5264           

Flag	Coverage Δ
#Debug	75.56% <0.00%> (?)
#production	71.50% <0.00%> (?)
#test	88.67% <ø> (?)

Impacted Files	Coverage Δ
...osoft.ML.Data/DataView/InternalSchemaDefinition.cs	55.40% <0.00%> (ø)
#Resolved

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@a2406f6). Click here to learn what that means.
The diff coverage is 77.20%.

@@            Coverage Diff            @@
##             master    #5114   +/-   ##
=========================================
  Coverage          ?   75.80%           
=========================================
  Files             ?      993           
  Lines             ?   181169           
  Branches          ?    19510           
=========================================
  Hits              ?   137330           
  Misses            ?    38535           
  Partials          ?     5304           

Flag	Coverage Δ
#Debug	75.80% <77.20%> (?)
#production	71.70% <75.22%> (?)
#test	88.92% <86.95%> (?)

Impacted Files	Coverage Δ
...osoft.ML.Data/DataView/InternalSchemaDefinition.cs	57.43% <50.00%> (ø)
src/Microsoft.ML.Data/Data/SchemaDefinition.cs	73.33% <74.46%> (ø)
src/Microsoft.ML.Data/DataView/TypedCursor.cs	82.53% <78.33%> (ø)
...oft.ML.Tests/OnnxSequenceTypeWithAttributesTest.cs	93.87% <86.95%> (ø)

[antoniovs1029]

/azp run #Resolved

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

#Resolved

[harishsk]

[antoniovs1029]

I've left some comments. Also, as discussed offline, I think it'd be useful to add a test replicating the original scenario of the original issue, catching the exception, and actually checking that the exception is now the one you've introduced. Thanks 😄