i#3048 func-trace: add drwrap flags, reduce func_trace instrument overhead

Created by: Louis-Ye

• Use DRWRAP_NO_FRILLS and DRWRAP_FAST_CLEANCALLS to make func_trace scale when instrumenting application with large number of threads.
• Use SAFE_READ* flags to prevent crashes.
• Remove operator new and delete from default option, since it is not stable in different architectures.
• Move consecutive markers append operation into one single function to reduce overhead of pre/post hook.
• Remove redzone checks.
• Update comments.

Issue #3048 (closed)