bench: aggregate adding completed ops for reads

[ivanvc]

Currently, the completed operations are added to the read benchmarks one by one, and given that each operation is atomic, it impacts the benchmark's performance. Change to update only once per cycle, with the total number of reads.

Related to #720

[ivanvc]

I debated using a defer or manually adding the call at the end of the loop. I ended up doing it with the defer, as there are some cases in which it would need to be called from inside the if condition if the run had an error and outside, too. So, it feels cleaner. However, I'm unsure if too many defers (one per loop) would create issues with the call stack if there are too many iterations.

[ahrtr]

Adding defer inside a for loop isn't a good practice.

How about this?

	err := func() error {
		defer func() { results.AddCompletedOps(numReads) }()

		c := tx.Bucket(benchBucketName).Cursor()
		for k, v := c.First(); k != nil; k, v = c.Next() {
			numReads++
			if v == nil {
				return ErrInvalidValue
			}
		}

		return nil
	}()
	if err != nil {
		return err
	}

[ahrtr]

Note that deferred function won't be called until the surrounding function returns. It means the initial motivation of progress notification is broken. My proposed change above can fix it.

[ivanvc]

@ahrtr, I applied the suggestion. PTAL :)

[ahrtr]

@fuweid @tjungblu PTAL, thanks

[tjungblu]

/lgtm

[ahrtr]

Also we can also do similar enhancement in runWritexxxx. Please also compare the write performance after you finish enhancing the runWritexxx. Thanks

[ivanvc]

Also we can also do similar enhancement in runWritexxxx. Please also compare the write performance after you finish enhancing the runWritexxx. Thanks

Hey @ahrtr, unfortunately (as I mentioned in #720 (comment)), changing the runWrite functions doesn't show an improvement. The results don't look consistent (I think it's similar to what @fuweid reported).

[ahrtr]

@ivanvc Please resolve the comment, also read #720 (comment)

[ivanvc]

@ivanvc Please resolve the comment, also read #720 (comment)

Sorry, @ahrtr, for some reason, I missed the notification and didn't read your comment. I'll address it later today.

[ahrtr]

lgtm

Thanks @ivanvc

cc @fuweid @tjungblu @Elbehery PTAL

[fuweid]

LGTM