tq/transfer_queue.go: fix retry delay trace log

In PR git-lfs#3449 we added support for the detection of 429 HTTP status code
responses to either Git LFS Batch API requests or individual object
transfer requests, with an implementation which expected a Retry-After
header with a specified retry time or delay period.  Commit
2197f76 of that PR introduced two
trace log output statements which attempted to report the delay
period calculated from the Retry-After header, one each in the
enqueueAndCollectRetriesFor() and handleTransferResult() methods of
the TransferQueue structure in our "tq" package.

Both of these statements attempted to format a value of the Duration
type (from the "time" package of the Go standard library) into a number
of seconds, expressed as a string, for their trace log messages.

However, the type returned by the Seconds() method of the Duration type
is a float64 type, so the trace log messages would contain format
error markers along with the floating-point value, for instance,
"%!s(float64=8.999544366) seconds".

The format string for one of these trace log output statements was
later updated in commit cf02216
of PR git-lfs#4097, when an exponential backoff technique was introduced to
better respect servers which sent responses with 429 status codes
but no Retry-After headers.  Specifically, the trace log message
in the enqueueAndCollectRetriesFor() method was revised to replace
the "%s" formatting verb for the float64 return type from the
Duration type's Seconds() method with a "%.2f" verb, which correctly
converts the float64 value into a string.

We now correct the related trace log message in the
handleTransferResult() to use the same "%2f" formatting verb for
the return type from the Duration type's Seconds() method.  We also
adjust the text of the message slightly to align more closely with the
corresponding message in the enqueueAndCollectRetriesFor() method.

Author: chrisd8088

tq/transfer_queue.go

@@ -806,7 +806,7 @@ func (q *TransferQueue) handleTransferResult(
 			// If the object can't be retried now, but can be
 			// after a certain period of time, send it to
 			// the retry channel with a time when it's ready.
-			tracerx.Printf("tq: retrying object %s after %s seconds.", oid, time.Until(readyTime).Seconds())
+			tracerx.Printf("tq: retrying object %s after %.2fs", oid, time.Until(readyTime).Seconds())
 			q.trMutex.Lock()
 			objects, ok := q.transfers[oid]
 			q.trMutex.Unlock()