What should provider's force_flush return

[cijothomas]

We can return an overall Result
Return a Vec of results from each processor/reader.

We need to be consistent across signals, unless there is a good reason to deviate.

originally discussed in : #1750 (comment)

[cijothomas]

From SIG meeting on 5/14

Make force_flush return Vec! of Results from each processor (logs,traces), reader (metrics).
The order in which processors are invoked are not guaranteed. (explicitly document this)
The error message should be standardized so it'll be possible to tell which processor/reader is failed.
Failing to flush one should not affect the rest.