Consider improving understandability of Exceptions (consumption)

**Is your feature request related to a problem? Please describe.**
Consuming models may raise exceptions which may be difficult to understand and pinpoint the issue.

For example, my case I reveice "Index was outside the bounds of the array." an then "Destination is too short"

In StackTrace I can see it it related ngrams, and thus, probably string values in my data. In this case, I was not able to find the reason yet even if exception happens regularly in some cases. It might be related to calling the model from multiple threads.

```
at Microsoft.ML.Data.NgramBufferBuilder.AddNgrams(VBuffer`1& src, Int32 icol, UInt32 keyMax)
at Microsoft.ML.Transforms.Text.NgramExtractingTransformer.Mapper.<>c__DisplayClass11_0.b__2(VBuffer`1& dst)
at Microsoft.ML.Transforms.LpNormNormalizingTransformer.Mapper.<>c__DisplayClass8_0.b__5(VBuffer`1& dst)
at Microsoft.ML.Data.ColumnConcatenatingTransformer.Mapper.BoundColumn.<>c__DisplayClass20_0`1.b__0(VBuffer`1& dst)
at Microsoft.ML.Data.ColumnConcatenatingTransformer.Mapper.BoundColumn.<>c__DisplayClass20_0`1.b__0(VBuffer`1& dst)
at Microsoft.ML.Data.SchemaBindablePredictorWrapperBase.<>c__DisplayClass19_0`2.b__0(TDst& dst)
at Microsoft.ML.Data.PredictedLabelScorerBase.EnsureCachedPosition[TScore](Int64& cachedPosition, TScore& score, DataViewRow boundRow, ValueGetter`1 scoreGetter)
at Microsoft.ML.Data.MulticlassClassificationScorer.<>c__DisplayClass16_0.b__0(UInt32& dst)
at Microsoft.ML.Transforms.KeyToValueMappingTransformer.Mapper.KeyToValueMap`2.<>c__DisplayClass8_0.b__0(TValue& dst)
at Microsoft.ML.Data.TypedCursorable`1.TypedRowBase.<>c__DisplayClass9_0`2.b__0(TRow row)
at Microsoft.ML.Data.TypedCursorable`1.TypedRowBase.FillValues(TRow row)
at Microsoft.ML.Data.TypedCursorable`1.RowImplementation.FillValues(TRow row)
at Microsoft.ML.PredictionEngineBase`2.FillValues(TDst prediction)
at Microsoft.ML.PredictionEngine`2.Predict(TSrc example, TDst& prediction)
at Microsoft.ML.PredictionEngineBase`2.Predict(TSrc example)
```
```

at System.ThrowHelper.ThrowArgumentException_DestinationTooShort()
at System.Span`1.CopyTo(Span`1 destination)
at Microsoft.ML.Data.BufferBuilder`1.GetResult(VBuffer`1& buffer)
at Microsoft.ML.Data.NgramBufferBuilder.GetResult(VBuffer`1& dst)
at Microsoft.ML.Transforms.Text.NgramExtractingTransformer.Mapper.<>c__DisplayClass11_0.b__2(VBuffer`1& dst)
at Microsoft.ML.Transforms.LpNormNormalizingTransformer.Mapper.<>c__DisplayClass8_0.b__5(VBuffer`1& dst)
at Microsoft.ML.Data.ColumnConcatenatingTransformer.Mapper.BoundColumn.<>c__DisplayClass20_0`1.b__0(VBuffer`1& dst)
at Microsoft.ML.Data.ColumnConcatenatingTransformer.Mapper.BoundColumn.<>c__DisplayClass20_0`1.b__0(VBuffer`1& dst)
at Microsoft.ML.Data.SchemaBindablePredictorWrapperBase.<>c__DisplayClass19_0`2.b__0(TDst& dst)
at Microsoft.ML.Data.PredictedLabelScorerBase.EnsureCachedPosition[TScore](Int64& cachedPosition, TScore& score, DataViewRow boundRow, ValueGetter`1 scoreGetter)
at Microsoft.ML.Data.MulticlassClassificationScorer.<>c__DisplayClass16_0.b__0(UInt32& dst)
at Microsoft.ML.Transforms.KeyToValueMappingTransformer.Mapper.KeyToValueMap`2.<>c__DisplayClass8_0.b__0(TValue& dst)
at Microsoft.ML.Data.TypedCursorable`1.TypedRowBase.<>c__DisplayClass9_0`2.b__0(TRow row)
at Microsoft.ML.Data.TypedCursorable`1.TypedRowBase.FillValues(TRow row)
at Microsoft.ML.Data.TypedCursorable`1.RowImplementation.FillValues(TRow row)
at Microsoft.ML.PredictionEngineBase`2.FillValues(TDst prediction)
at Microsoft.ML.PredictionEngine`2.Predict(TSrc example, TDst& prediction)
at Microsoft.ML.PredictionEngineBase`2.Predict(TSrc example)
```

**Describe the solution you'd like**
Unsure. Maybe showing, something like "Column XXX throwed an exception" and then putting the actual Exception inside the InnerException?

Or, maybe Model Builder could include a "Create debuggable model" which would add the necessary source code the debug such issues without manual setup?

Or, a guide on how to enable debuggability manually, such as by downloading Microsoft.ML source code and referencing directly from local project?

**Describe alternatives you've considered**
I might try adding Microsoft.ML source code directly to my project and reference it from there. Are there any known issues doing this, such as version mismatch?

Le tracce dello stack del report identificano NgramBufferBuilder.AddNgrams, BufferBuilder.GetResult e PredictionEngine.Predict come punti di ingresso; iniziate tracciando il modo in cui le eccezioni si propagano attraverso queste chiamate. Per considerare il lavoro completato, sarebbe necessario concordare un metodo riproducibile per aggiungere informazioni sulla colonna o sul contesto alle eccezioni rivolte ai consumer, con la validazione dell'n-gram segnalato e degli scenari di uso concorrente.

Scritto dal modello di indicizzazione a partire dal testo della issue.