[SharedCache] Optimize parsing and applying of slide info

[bdash]

Slide info is parsed and applied on the main thread, below the SharedCache constructor, when a shared cache is opened. Time spent applying the slide is time that the main thread is blocked. This collection of optimizations cuts the time spent applying the slide roughly in half (125ms to 65ms on a macOS shared cache).

The individual changes are:

1. Avoid building the rewrites vector unless SLIDEINFO_DEBUG_TAGS is enabled. Building the vector was the single most expensive operation in the function.
2. Read slide pointers via ReadULong rather than the variable-length Read method. The compiler wasn't able to eliminate the call to memcpy in the variable-length Read function, and the overhead of calling memcpy is non-trivial given the small size of the read and number of times it is called.
3. Update ParseAndApplySlideInfoForFile to take a pointer rather than a std::shared_ptr. This method doesn't need to manipulate the ownership of the object so the shared_ptr is unnecessary.

After these changes the profile shows 80% the time spent below ParseAndApplySlideInfoForFile is inside MappedFileAccessor::ReadULong / MappedFileAccessor::WritePointer.