Build errors with GCC + ASAN/UBSAN

[ravyu-jump]

Building with GCC + ASAN / UBSAN appears to treat certain checks as errors.

With ASAN, we have -Werror=stringop-truncation errors like so:

src/app/fdctl/config.c: In function 'fdctl_cfg_from_env':
src/app/fdctl/config.c:473:5: error: '__builtin_strncpy' output may be truncated copying 15 bytes from a string of length 255 [-Werror=stringop-truncation]
  473 |     strncpy( config->tiles.net.interface,
      |     ^
src/app/fdctl/config.c:490:3: error: '__builtin_strncpy' output may be truncated copying 39 bytes from a string of length 64 [-Werror=stringop-truncation]
  490 |   strncpy( config->hostname, utsname.nodename, sizeof(config->hostname) );
      |   ^
cc1: all warnings being treated as errors
make: *** [config/everything.mk:308: build/test-vectors/obj/app/fdctl/config.o] Error 1

UBSAN sees sign-conversion and stringop-overflow checks triggered like so:

src/ballet/base58/../../util/spad/../bits/fd_bits.h:681:50: error: writing 4 bytes into a region of size 0 [-Werror=stringop-overflow=]
  681 |   (__extension__({ T _fd_store_tmp = (val); (T *)memcpy( (T *)(dst), &_fd_store_tmp, sizeof(T) ); }))
      |                                                  ^~~~~~
src/ballet/base58/fd_base58_tmpl.c:380:5: note: in expansion of macro 'FD_STORE'
  380 |     FD_STORE( uint, &out_as_uint[ i ], fd_uint_bswap( (uint)binary[ i ] ) );
      |     ^~~~~~~~

src/ballet/pack/fd_pack.c: In function 'fd_pack_clear_all':
src/ballet/pack/fd_pack.c:817:21: error: conversion to 'ulong' {aka 'long unsigned int'} from 'int' may change the sign of the result [-Werror=sign-conversion]
  817 |       ulong __idx = (idx);                                                                     \
      |                     ^
src/ballet/pack/fd_pack.c:1783:38: note: in expansion of macro 'ACCT_IDX_TO_PTR'
 1783 |       fd_acct_addr_t penalty_acct = *ACCT_IDX_TO_PTR( FD_ORD_TXN_ROOT_PENALTY_ACCT_IDX( del->root ) );
      |                                      ^~~~~~~~~~~~~~~
src/ballet/pack/fd_pack.c: In function 'fd_pack_delete_transaction':
src/ballet/pack/fd_pack.c:817:21: error: conversion to 'ulong' {aka 'long unsigned int'} from 'int' may change the sign of the result [-Werror=sign-conversion]
  817 |       ulong __idx = (idx);                                                                     \
      |                     ^
src/ballet/pack/fd_pack.c:1834:38: note: in expansion of macro 'ACCT_IDX_TO_PTR'
 1834 |       fd_acct_addr_t penalty_acct = *ACCT_IDX_TO_PTR( FD_ORD_TXN_ROOT_PENALTY_ACCT_IDX( root_idx ) );

[ripatel-fd]

I looked at a couple of them and they seem to be compiler bugs.

It seems like GCC warnings change unpredictably with every major release and certain compile flags.
stringop-overflow, stringop-truncation, and maybe-uninitialized produce frequent false positives.

Maybe we should move these unreliable checks to with-extra-brutality.mk so they are not enabled by default.