marking data to be delayed if the previous pending is inflight

[wei-lee]

To reproduce the issue:

• Use Charles to inspect the network calls
• start the client app
• stop the cloud app
• create a new record in the client. Then immediately do 2 updates to the record
• start the client app again.
• When sync loop starts, you will see in the request body, all the changes are sent at once.
• The app may display old value for the record.

The fix will make sure if a record is pending and inflight, the subsequent changes to that record will be marked as delayed and will not be sent until the pending change is resolved.

[corinnekrych]

@wei-lee To avoid crash on start-up exception due to https://github.com/feedhenry/fh-ios-sdk/pull/50/files#diff-098e1ff33bf2d868069cb8aa5d075d76R180, i locally added this line.

Doing the test "Update while CloudApp is crashed Update when restarted" See https://docs.google.com/spreadsheets/d/1bMY8QqfE9SD_xRpBBPFgj-byqxPYFmIjHR1ig6Mxq7c/edit#gid=1130697588 column A with code on Master column B with code on this PR

I still have data update loses and i can not send update any more. @wei-lee I don't understand how "crash" field is supposed to be used in the server/client sync protocol.