Array I/O support for julia-1.6

[fhagemann]

In the current implementation of I/O of arrays, the number of dimensions cannot be properly checked in julia-1.6 because

ndims(AbstractArray{<:RealQuantity, N})

is not defined (julia-1.6 does not like the <: in the first argument.

Therefore, ndims was replaced by _ndims which allows for handling this special case for array I/O.

[codecov[bot]]

Codecov Report

Patch coverage: 100.00% and project coverage change: +4.53% :tada:

Comparison is base (3176874) 48.31% compared to head (da12a9a) 52.85%.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #33 +/- ## ========================================== + Coverage 48.31% 52.85% +4.53% ========================================== Files 6 6 Lines 594 596 +2 ========================================== + Hits 287 315 +28 + Misses 307 281 -26 ``` | [Files Changed](https://app.codecov.io/gh/legend-exp/LegendHDF5IO.jl/pull/33?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=legend-exp) | Coverage Δ | | |---|---|---| | [src/LegendHDF5IO.jl](https://app.codecov.io/gh/legend-exp/LegendHDF5IO.jl/pull/33?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=legend-exp#diff-c3JjL0xlZ2VuZEhERjVJTy5qbA==) | `100.00% <ø> (ø)` | | | [src/generic\_io.jl](https://app.codecov.io/gh/legend-exp/LegendHDF5IO.jl/pull/33?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=legend-exp#diff-c3JjL2dlbmVyaWNfaW8uamw=) | `41.31% <100.00%> (+9.50%)` | :arrow_up: | | [src/types.jl](https://app.codecov.io/gh/legend-exp/LegendHDF5IO.jl/pull/33?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=legend-exp#diff-c3JjL3R5cGVzLmps) | `90.16% <100.00%> (ø)` | |

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[fhagemann]

Sorry for all the force pushes, I initially made very stupid mistakes.. The nightly tests fail but this seems unrelated --> this is ready to be merged now @oschulz

If we released a new version v0.1.6 after merging, then we'd also allow for StatsBase 0.34 without requiring to load the main branch.