Enhance logging infrastructure to cope with C and C++ style strings.

[AlexHockey]

This PR enhances our logging infrastructure to allow you to pass in C++ strings. This has two advantages:

• You don't have to tediously type .c_str() everywhere.
• You avoid bugs where you accidentally do things like TRC_DEBUG("Hello %s", get_name().c_str()); and call .c_str() on an rvalue, which results in a dangling pointer.

I've tested this by changing one of the logs in Ralf to pass in a C++ string, and checked that the log still works.

I've wrote this code and disassambled it.

std::string name() { return"Alex"; }
voidtest() { TRC_DEBUG("Hello %s", name()); }

The test function calls the correct template instantiation:

 0x00000000004615af <+281>:	callq 0x4649b3 <Log::write<std::string>(int, char const*, int, char const*, std::string)>

which calls the right instantiation of _loggable:

 0x00000000004649e6 <+51>:	callq 0x465d47 <Log::_loggable<char const*>(std::string const&)>

which does call c_str 🙂

 0x0000000000465d47 <+0>:	push %rbp
0x0000000000465d48 <+1>:	mov %rsp,%rbp
0x0000000000465d4b <+4>:	sub $0x10,%rsp
0x0000000000465d4f <+8>:	mov %rdi,-0x8(%rbp)
0x0000000000465d53 <+12>:	mov 0x82ac86(%rip),%rax # 0xc909e0 <__gcov0._ZN3Log9_loggableIPKcEES2_RKSs>
0x0000000000465d5a <+19>:	add $0x1,%rax
0x0000000000465d5e <+23>:	mov %rax,0x82ac7b(%rip) # 0xc909e0 <__gcov0._ZN3Log9_loggableIPKcEES2_RKSs>
0x0000000000465d65 <+30>:	mov -0x8(%rbp),%rax
0x0000000000465d69 <+34>:	mov %rax,%rdi
0x0000000000465d6c <+37>:	callq 0x4133d0 <_ZNKSs5c_strEv@plt>
0x0000000000465d71 <+42>:	mov 0x82ac70(%rip),%rdx # 0xc909e8 <__gcov0._ZN3Log9_loggableIPKcEES2_RKSs+8>
0x0000000000465d78 <+49>:	add $0x1,%rdx
0x0000000000465d7c <+53>:	mov %rdx,0x82ac65(%rip) # 0xc909e8 <__gcov0._ZN3Log9_loggableIPKcEES2_RKSs+8>
0x0000000000465d83 <+60>:	leaveq 0x0000000000465d84 <+61>:	retq 

This also confirms that the rvalue that is returned by name() is converted to an lvalue by the time that the function that actually does the logging is called.

I was also worried about binary sizes. I checked building sprout with and without this change. The sprout executable was 79MB before and 80MB after, so I'm not too worried.