_read_results gives more data than requested when using a larger cache

[GabrielKS]

The symptom is kind of the opposite as in #1072 — we get more timestamps than called for when the cache contains a superset in time of what we need:

julia> timestamps = PSI._process_timestamps(res, DateTime("2035-01-01"), 7);

julia> variable_tuples = [(ActivePowerVariable, ThermalStandard)];

julia> variable_keys = [PSI.VariableKey(ActivePowerVariable, ThermalStandard)];

julia> load_results!(res, 7; variables=variable_tuples)

julia> PSI._read_results(res, variable_keys, timestamps[1:6], nothing) |> values |> first |> x->x.data |> keys |> collect
7-element Vector{DateTime}:
 2035-01-01T00:00:00
 2035-01-02T00:00:00
 2035-01-03T00:00:00
 2035-01-04T00:00:00
 2035-01-05T00:00:00
 2035-01-06T00:00:00
 2035-01-07T00:00:00

julia> # 7 timestamps in the result even though we asked for only 6

This causes problems downstream when we convert to DataFrame, see #1073 (comment).

[GabrielKS]

Fixed in psy4