Net::HTTP has default infinite timeout for opening connection, which,
when other side does not open connection fully, nor rejects connecting,
results in infinite thread hang without any reporting into log.

Fix it by explicitly making all timeouts finite (set to default
Net::HTTP's read_timeout 60 seconds).

/cc @klaus

Previously WendelinClient's tracing was controlled by global $trace
variable and was not correlated to current Fluentd log level, or current
plugin log level and thus it was harder for users to use.

Improve it and rework the tracing to use fluentd logging infrastructure.

This way tracing is disabled by default, but can be either globally
enabled with

    fluentd -v -v

or only for wendelin_out plugin via adding

    log_level   trace

to <match ...> section which routes messages to wendelin_out.

/cc @klaus

Instead of ingesting just [1,2,'hello'] list, ingest in format closer to what
fluentd sends - i.e. "t" and dict/hash of name->value pairs.

This was pre-tested and know to ingest data ok, modulo some known TODO.

The project to implement Fluentd output to Wendelin.