Use epicsTypes.h for conditions related to Base version

[ralphlange]

Following a suggestion from @mdavidsaver, here is a better fix for #44.

[mdavidsaver]

Looks like progress!

[mdavidsaver]

I notice another bit of weirdness. The VALUE union definition is also conditional on DBR_INT64, but does not directly include any header which defines it. To make qtcreator shut up about this I tried adding #include <dbAccess.h> to caPutLogTask.h. This led me to realize that caPutJsonLogTest.cpp includes db_access.h, which does not defined DBR_INT64. So different files within caPutLog have different views of this union. Not practically a problem since the double member has the same size and (I think) the same or stricter alignment. Still, exposing this union seems a magnet for ABI problems.

[ralphlange]

Given that this was contractor work, one would wonder how thoroughly it was reviewed...

[ralphlange]

I opened another ticket - see above.

Hiding this union, which I support, would require review and changes in the CA Gateway interface to caPutLog. Needs coordinated PRs etc. etc.

[mdavidsaver]

Or alternately, since this PR appears to remove base 3.14 support, adjust the yajl usage to work back to 3.15 and make JSON support unconditional. Also, I think it would be possible to restructure so that caPutJsonLogTest does not see union VALUE, which does not seem to be directly used there. This way union VALUE could be tied to dbAccess.h.

[ralphlange]

... since this PR appears to remove base 3.14 support, ...

As far as I can see, 3.14 support has been officially removed with R4.1, by using epicsAtomic.h. This PR is just following up.

[ralphlange]

Superseded by #50