Update Filesystem Benchmarking to be more strict

Early results indicated that 10-20 second benchmarks on FS writes would be acceptable but not the best. In the last few weeks we've seen a couple of customers that had just over 10 second results from this test and had horrible GitLab performance.

As a result, let's tighten up our benchmarks and add a note about the naivety of this test. Since local SSD can write these files in less than 1 second, allowing up to 10 seconds for 'best' performance seems like a relatively huge range, still.

Author's checklist

• Apply the correct labels and milestone
• Crosslink the document from the higher-level index
• Crosslink the document from other subject-related docs
• Feature moving tiers? Make sure the change is also reflected in [features.yml](https://gitlab.com/g tlab-com/www-gitlab-com/blob/master/data/features.yml)
• Correctly apply the product [badges](https://docs.gitlab.com/ee/development/docum ntation/styleguide.html#product-badges) and [tiers](https://docs.gitlab.com/ee/development/documentation/styleguide.ht l#gitlab-versions-and-tiers)
• Port the MR to EE (or backport from CE): always recommended, required when the ee-compat-check job fails

Review checklist

• Your team's review (required)
• PM's review (recommended, but not a blocker)
• Technical writer's review (required)
• Merge the EE-MR first, CE-MR afterwards