LRUCacheShard cache line size alignment

[grooverdan]

combining #2568 and #2612.

[grooverdan]

mingw32 failure - https://gcc.gnu.org/bugzilla/show_bug.cgi?id=52991

[facebook-github-bot]

@IslamAbdelRahman has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[IslamAbdelRahman]

Thanks @grooverdan, How can we suppress the MinGW failure .. we don't want our travis failure to consistently fail

[grooverdan]

I tried just adding -mno-ms-bitfields on the mingw32 however this wasn't sufficient https://travis-ci.org/grooverdan/rocksdb/jobs/255882007

So as per next commit just disabling alignment for mingw32

[facebook-github-bot]

@grooverdan updated the pull request - view changes - changes since last import

[facebook-github-bot]

@IslamAbdelRahman has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

[IslamAbdelRahman]

Thanks!

[grooverdan]

Thanks for the initial work and merging.

[yiwu-arbug]

Seems this still doesn't properly align LRUCacheShard, as shown by running tests with UBSAN: https://gist.github.com/yiwu-arbug/401217b8868be158a545f85220785e17

I guess it is because we allocate shard_ as an array, which short-circuit the overloaded new operator.