Note that srun reset this only once, while salloc/sbatch reset
each pass. Go with salloc/sbatch behavior - I do not think
carrying the distribution over between components makes sense
for most allocations at this time.

Note this overlaps with the --exclusive option, but inverts
the default behavior if no arguments are given.

Note that srun was previously using NO_VAL as unset which is technically
an acceptable value; change it to -1 to match salloc/sbatch.

Note changes needed for both the environment variable handling, as well
as for PBS option parsing; have them call in to slurm_process_option().

Simplify error handling, and explicitly test for NO_VAL as a return from
time_str2mins().

There are two equivalent long-form options. Add a new LONG_OPT_CLUSTER
to the enum to distinguish which form is used. (Needed for tracing at
a later point in time.)

Neither salloc/sbatch reset this. I believe reseting each pass,
as is done by srun, is the correct behavior, otherwise each separate
component will request additional licenses which is unintuitive.

For reset, salloc always reset this where sbatch/srun would not.
Choosing to go with sbatch/srun's behavior.

This should be avoided, but if there are multiple slurm_opt_t entries for
a given val (e.g., -W in sbatch vs srun) check that we've found the correct
entry.

Non-single-character options do not have this issue since they should just
use differing LONG_OPT values.

A number of validation routines - especially within srun -
care whether the variable was set through the environment, or by
an explicit CLI option (or embedded in a job script).

Provide an easy central test so this can be handled for all arguments.

The envvar interface requires some translation to be handled
in the same manner of options returned from getopt_long(),
as it will intentionally return empty strings for certain
variable types.

Match OPT_BOOL processing for flags (defined as has_arg = 0 for
their getopt_long equivalents).