double-typed buff_len

[cboulay]

This adds new methods to the C API and an overloaded method to the C++ API to allow the user to set the buffer length with a double instead of an integer. This is necessary in very-high channel count and samplerate situations where even 1-second is too much. Obviously a sub-1-second buffer means that the system won't be able to recover from a disconnect without data loss but at these data rates that isn't really possible anyway (unless we want to start buffering to disk).

[cboulay]

@tstenner had a good idea to make this lsl_create_outlet_ex and lsl_create_inlet_ex, where the new additional argument can receive bitwise-OR'd flags. So in common.h, we'd have

typedef enum {
    transp_default = 0,                        // keep legacy behavior
    transp_buffsize_samples = 1,       // the supplied int is in samples
    transp_buffsize_thousandths = 2, // Indicates whether the supplied int should be (double)/1000
    transp_synch = 4,                          // the outlet will use synchronous (blocking) zero-copy pushes
    transp_disable_endianness = 8,   // disable endianness conversions
} lsl_transport_options_t;

[cboulay]

This is such a trivial detail, but do we have any consistency as to whether the common short-form for 'buffer' should be buf or buff?

[cboulay]

Rebased on #155. Let's wait on that one to be merged so we can easily see the (comparatively very small) changes.

[tstenner]

Rebased and merged in 0b35dba3464c22bc707fef88d18483c597519e81