Drop separate table files keeping only merged ones. Add tables for MF model 170812

[namapane]

Merged MF interpolation files have been introduced in cms-sw/cmssw#38051, but at the time the merged files were added only for the 3.8 T working point, and without removing the separate files to avoid breaking unit tests.

With cms-sw/cmssw#51526, the mechanism to use merged files has been extended to all MF builders, so that it is now possible to stop distributing thousands of individual files and keep only the merged ones.

In this PR, the 11136 individual data files are removed for each of the 4 160812 maps (3.8 T, 3.5 T, 3 T, 3.8 T@Run I), replacing them by the merged files when not already present.

In addition, merged tables for all nominal working points (including 2 T, which was missing in 160812) for the "legacy" map version 170812 are added. Compared to 160812, this model mainly includes corrected steel magnetization curves for ss400 steel [cf. here]. The resulting difference was found to be very small, and for this reason at the time consensus was not reached to release this model.
However, I believe it is important to release these tables for the long term future of CMS because:

• This is the only set of tables that is complete with a consistent model for all nominal working points. It may become important if, for whatever reason in the future of CMS, it is decided to operate the magnet at different working points than 3.8 T.
• This model includes the ultimate knowledge of the CMS magnet and yoke; no further development is planned. It is not granted that expertise will still be available in a timescale of a decade, so better preserve this work for good.
• Access to this model was requested by proponents of the W mass analysis for comparison with historical NMR data and for systematic studies [link]. This has been provided with ad-hoc setups, but other use cases may arise in the future.

Notes

• Nothing changes in the default maps. The 170812 version will be available customizing the standard sequences. (Switching to 170812 as the default for Run4 and beyond makes sense in my opinion but it's not the purpose of this PR).
• This PR requires CMSSW versions which include Updates and maintenance of the MagneticField subsystem cms-sw/cmssw#51526, i.e. current 20_1_X integration builds, or the upcoming 20_1_0_pre3 when it will be released.

Testing

Tested the merged tables for all 160812 models with the MF regression (checks for identical results over a large number of random points) with all MF builders ((static, from DB) x (DDD, dd4HEP))
Tested the 170812 models for the expected differences in the same regression.

[namapane]

@perrotta: FYI