…on-error

Use const char* for requestType.

Overhauled README

uses char instead of int

credit to https://github.com/salanki

…some HTTP implementations

For example: Thin for Ruby users. Credit to https://github.com/salanki

Adjust Content-Length and data (only one \n at the end

Adjust line delimiter, correct end of headres search

…de on invalid returns

Added error handling to skipHeaders to prevent usable of httpReturnCode
on invalid returns

Simplified check because it is only relevant if the status code could
be read - all other error cases are handled by later code too

When the server closed the connection and all data are read, the socket
is in „CLOSE-WAIT“ status … but in this combination „connected“ returns
false, even if the socket is not really closed on the client.

This is not a real problem because normally „CLOSE_WAIT“ sockets are
reused for new connections, but it may end up in problems any may cause
problems. (I try to tackle down the effect that sometimes no more
connections are possible …)

The solution is to call „stop“ in any case when the Stream should be
closed because with this the connection is closed correctly in any
case. A double call is handed by „stop“ automatically so don’t worry
about this.

- Added some checks that stream variable is not NULL before using it
(just to make sure)
- To avoid header-read errors that happened sometimes implement a check
for available data with a maximum timeout of 1000ms (configurable via
defines in header file). With my default settings we wait 20 ms and
then check again if at least 17 characters are available to read (which
should be a normal HTTP-header first line). This is done at max 50
times before the read is tried

Added error handling to skipHeaders to prevent usable of httpReturnCode on invalid returns