bigquery: add missing_value_interpretations as append option

[brachipa]

As describe here

The Storage Write API only populates default values when the write stream schema is missing a field that is contained in the destination table schema

So when we want to default value we need to make sure the field is missing from the schema, what makes our code ugly, we need to iterate over the schema fields and delete the column with the default value (field_with_default_value)

	md, err := bqClient.Dataset(datasetName).Table(tableName).Metadata(ctx)
	if err != nil {
		...
	}

	schema := md.Schema
	convertedSchema, err := adapt.BQSchemaToStorageTableSchema(schema)
	if err != nil {
		...
	}
	savedInd := -1
	for i, field := range convertedSchema.Fields {
		if field.Name == "field_with_default_value" {
			savedInd = i
			break
		}
	}
	if savedI > 0 {
		convertedSchema.Fields = append(convertedSchema.Fields[:savedInd], convertedSchema.Fields[savedInd+1:]...)
	}

Or:

You can also specify request-level default values in the missing_value_interpretations map within the AppendRowsRequest message. Each key is the name of a column and its value indicates how to interpret missing values.
For example, the map {'col1': NULL_VALUE, 'col2': DEFAULT_VALUE} means that all missing values in col1 are interpreted as NULL, and all missing values in col2 are interpreted as the default value set for col2 in the table schema.

Unfortunately, we have no way to pass this map into the append request!

The solution:

Add new option WithMissingValueInterpretations to pass this information

missingValueInterpretation := map[string]storagepb.AppendRowsRequest_MissingValueInterpretation{
"field_with_default_value": storagepb.AppendRowsRequest_DEFAULT_VALUE}

....

result, err = stream.AppendRows(ctx, [][]byte{b}, managedwriter.WithMissingValueInterpretations(missingValueInterpretation))

[shollyman]

Thanks for the request.

Currently the missing_value_interpretations is per-request in the underlying API. Would you expect the option in managedwriter to behave more like schema, which persists across requests, as a per-append option?

[brachipa]

@shollyman yes, per request is ok.
I already submit a pr with a fix.
#8388