Implement `UnitConverter` from/to units in base `Transform` class

[StokesMIDE]

One of the main functions of the ‘calibration’ polynomials is converting from the device’s internal units to actual engineering units. It would make sense for the polynomials themselves to be able to specify the units (and/or measurement type) they output. If defined, these would override any units defined by the object (Channel, SubChannel, or EventArray) using it. UnitConverter implements this. This code could be moved from there to the base Transform class, and the related code in EventArray would be moved into the base/mixin Transformable class.

For the present, this would not have any outward effect on idelib; current Transforms (other than the UnitConverters) wouldn't have these attributes set, so the units would fall back to SubChannel's definition. However, there is discussion on the enDAQ firmware side about writing this information in the IDE's CalibrationList; having it set up on the idelib side in advance would make it easier.

[StokesMIDE]

These are the proposed EBML elements:

<UnicodeElement name="PolynomialOutputUnits" id="0x4B30" multiple="0"/>
<UnicodeElement name="PolynomialOutputUnitsName" id="0x4B31" multiple="0"/>
<UIntegerElement name="PolynomialOutputUnitsXRef" id="0x4B32" multiple="0"/>
<UnicodeElement name="PolynomialInputUnits" id="0x4B40" multiple="0"/>
<UnicodeElement name="PolynomialInputUnitsName" id="0x4B41" multiple="0"/>
<UIntegerElement name="PolynomialInputUnitsXRef" id="0x4B42" multiple="0"/>

Notes:

• Since compatibility is determined by the units, the <PolynomialInputUnitsName> may not be necessary, but we might as well allocate an ID for it (might be used for UI purposes, like a menu item 'Display x as y').
• The “XRef” elements match the existing (but unused) <SubChannelUnitsXRef> element, which is intended for localization. Even if we currently have no plans to use it, we should allocate it an ID for compatibility with <SubChannel>.
• We could reuse the SubChannelUnits* elements in the polynomials for output units; it would remove some redundancy and save an iota of memory (and schema ID space), but the names would be confusing. Then again, users wouldn’t see these names unless they were looking at raw EBML data.