guest: clean up and unify reading the timestamp from a variable buffer

[erichte-ibm]

In multiple places in the code (namely guest_svc_read.c:read_path()), we load in data from a variable, then proceed to do some level of validation which mostly consists of checking that the buffer is an expected size before doing any parsing. As asked in PR #75 , the "expected" size of a buffer should be defined, documented, and probably abstracted out so that other places of the code can call a function to read in a particular variable and do this same validation.

---

From my understanding, the following conditions should be checked and handled on the buffer size:

• size == 0 → error case, likely invalid file or some other I/O error. variables should not be completely empty
• 0 < size < HEADER_LEN → error case, variable must contain at least the 16-byte header
• size == HEADER_LEN → valid case, variable exists but contains no data
• size > HEADER_LEN → valid case, data exists and can be parsed

Data read from a variable contains a header of 16 bytes, consisting of a 1-byte version number, then 15 bytes of timestamp, truncating the last pad so that the header remains aligned.

---

Cleaning this up will probably looking some like:

• factor out header parsing into another function, perhaps variable_from_path()
• remove references to TIMESTAMP_LEN and TIMESTAMP_LEN - 1 when referring to this header
• actually parse and inspect the version number?
• clean up any duplicate code paths that differentiated reading from a variable vs reading from a local file

[erichte-ibm]

Resolved in #77