Fix follow_exec latent problem

Fix follow_exec latent problem

A following commit to make each inferior have its own thread list
exposes a problem with bf93d7ba99 ("Add thread after updating gdbarch
when exec'ing"), which is that we can't defer adding the thread
because that breaks try_open_exec_file which deep inside ends up
calling inferior_thread():

 #5  0x0000000000637c78 in internal_error(char const*, int, char const*, ...) (file=0xc151f8 "src/gdb/thread.c", line=165, fmt=0xc15180 "%s: Assertion `%s' failed.") at src/gdb/common/errors.c:55
 #6  0x00000000008a3d80 in inferior_thread() () at src/gdb/thread.c:165
 #7  0x0000000000456f91 in try_thread_db_load_1(thread_db_info*) (info=0x277eb00) at src/gdb/linux-thread-db.c:830
 #8  0x0000000000457554 in try_thread_db_load(char const*, int) (library=0xb01a4f "libthread_db.so.1", check_auto_load_safe=0)
     at src/gdb/linux-thread-db.c:1002
 #9  0x0000000000457861 in try_thread_db_load_from_sdir() () at src/gdb/linux-thread-db.c:1079
 #10 0x0000000000457b72 in thread_db_load_search() () at src/gdb/linux-thread-db.c:1134
 #11 0x0000000000457d29 in thread_db_load() () at src/gdb/linux-thread-db.c:1192
 #12 0x0000000000457e51 in check_for_thread_db() () at src/gdb/linux-thread-db.c:1244
 #13 0x0000000000457ed2 in thread_db_new_objfile(objfile*) (objfile=0x270ff60) at src/gdb/linux-thread-db.c:1273
 #14 0x000000000045a92e in std::_Function_handler<void (objfile*), void (*)(objfile*)>::_M_invoke(std::_Any_data const&, objfile*&&) (__functor=..., __args#0=@0x7ffef3efe140: 0x270ff60) at /usr/include/c++/7/bits/std_function.h:316
 #15 0x00000000007bbebf in std::function<void (objfile*)>::operator()(objfile*) const (this=0x24e1d18, __args#0=0x270ff60)
     at /usr/include/c++/7/bits/std_function.h:706
 #16 0x00000000007bba86 in gdb::observers::observable<objfile*>::notify(objfile*) const (this=0x117ce80 <gdb::observers::new_objfile>, args#0=0x270ff60) at src/gdb/common/observable.h:106
 #17 0x0000000000856000 in symbol_file_add_with_addrs(bfd*, char const*, symfile_add_flags, section_addr_info*, objfile_flags, objfile*) (abfd=0x1d7dae0, name=0x254bfc0 "/ho

The problem is latent currently because inferior_thread() at that
point manages to return a thread, even though it's the wrong one (of
the old inferior).

The problem originally fixed by bf93d7ba99 was:

    (...) we should avoid doing register reads
    after a process does an exec and before we've updated that inferior's
    gdbarch.  Otherwise, we may interpret the registers using the wrong
    architecture.

    (...) The call to "add_thread" done just after adding the inferior is
    problematic, because it ends up reading the registers (because the ptid
    is re-used, we end up doing a switch_to_thread to it, which tries to
    update stop_pc). (...)

The register-reading issue is no longer a problem nowadays, ever since
switch_to_thread stopped reading the stop_pc in git commit
f2ffa92bbce9 ("gdb: Eliminate the 'stop_pc' global").

So this commit basically reverts bf93d7ba99.

gdb/ChangeLog:
2018-11-22  Pedro Alves  <palves@redhat.com>

	* infrun.c (follow_exec) <set follow-exec new>: Add thread and
	switch to it before calling into try_open_exec_file.