fixes two bugs in the Memory module

[ivg]

The first bug was breaking the raw loader at least. It bug was
introduced in #1178 during the Memory.view optimization. A check was
missed that allowed for creation of invalid views. The invalid views
later were caught with an assertion check, (easily reproduced with
bap /bin/ls --loader=raw), e.g.,

("Assert_failure bap_memory.ml:72:2"
  "Raised at file \"bap_memory.ml\", line 72, characters 2-31\
 \nCalled from file \"bap_memory.ml\", line 196, characters 2-18\
 \nCalled from file \"bap_memory.ml\", line 410, characters 26-52\
 \nCalled from file \"bap_trie.ml\", line 46, characters 34-53\
 \nCalled from file \"bap_trie.ml\", line 62, characters 4-130\
 \nCalled from file \"bap_byteweight.ml\", line 78, characters 12-39\
 \nCalled from file \"bap_byteweight.ml\", line 129, characters 12-39\
 \nCalled from file \"byteweight_main.ml\", line 50, characters 44-56\
 \nCalled from file \"src/sequence.ml\", line 123, characters 29-36\
 \nCalled from file \"byteweight_main.ml\" (inlined), line 49, characters 4-116\
 \nCalled from file \"byteweight_main.ml\", line 48, characters 4-146\
...

The second bug was probably all the time in the library and concerns
Memory.find_map and Memory.find_if, which are talking an optional
word_size parameter that was ignored and functions were always
iterating over bytes.