[output.wavefront] Introduced "immediate_flush" flag

[prydin]

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

Addresses an issue mentioned in #7409 concerning the "double buffering" done by the Wavefront SDK. The SDK is designed to send metrics asynchronously in order to minimize the delay experienced by a caller. This behavior is not ideal when used with Telegraf, since Telegraf will assume the connection to Wavefront can accept data very quickly. When the internal buffer in the Wavefront SDK overflows, it will drop metrics. Meanwhile, Telegraf will still think the metrics are flowing at a very high rate and will keep flooding the SDK.

To avoid extensive redesign of the SDK, this PR adds a flag, immediate_flush that flushes the internal SDK buffers after each batch. This will ensure that the buffer never fills with more records than the batch size and also has the side effect of providing "back pressure" to the Telegraf core. Thanks to that back pressure, Telegraf will no longer be "tricked" into thinking the connection has virtually unlimited capacity and can take whatever the action it needs to manage the throughput gracefully.

[ssoroka]

This is great, but I think there's no reason to still support the buffer or allow configuration of the flush. Keep in mind Telegraf is already doing periodic flush (which is essentially the write method) based on metric_batch_size and flush_interval.