Grok Record Reader applies regex to each line twice in some situations

When using the Grok Record Reader, it reads the next line from the stream to check if it needs to be appended to the record, is the start of a stack trace, etc. When this is done, if the next line matches the configured pattern, it is stored as 'nextLine' but we don't store the Map that is obtained from calling captures(). This means that the next iteration of nextRecord() must re-evaluate the regexes against 'nextLine'. We should instead just keep the Map as a member variable.