Address hardcoded units MACRO calibration and M-M linkage

[khaeru]

To complete the code added #223, address the point raised there by @gidden:

There are a variety of places in the previous R code as well as existing GAMS code where unit conversions are hardcoded (usually going from $B to $T via either / 1000 or / 1e3).

(To clarify: message_ix.macro code was transcribed from an R script. The unit conversions were carried into the Python code. This is undesirable since the code should operate with whichever units the user inputs.)

This can be resolved in one or more of the following ways:

1. At a minimum, clearly describe (e.g. in message_ix.macro lacks minimal documentation #315) the meaning/difference between units of quantities in MACRO and MESSAGE, so that users can understand how to interpret their values.
2. Per TODOs inline in message_ix.macro: read, parse, and follow the existing units in the Scenario being modified.
3. Remove hard-coded conversions in the MESSAGE or MACRO GAMS code.

[khaeru]

Postponing to after v3.0.

[khaeru]

No one was assigned to do this by the feature freeze for v3.2.0, so we postpone to v3.3 (or later).

[khaeru]

No one was assigned to do this by the feature freeze for v3.3.0, so we postpone to v3.4 (or later).

[LauWien]

No one was assigned to do this for v3.4.0, so we postpone to v3.5 (or later).

[LauWien]

No one was assigned to do this by the feature freeze for v3.5.0, so we postpone to v3.6 (or later).

[khaeru]

Removing the milestone entirely. It can be re-added if/when the team decides to schedule/assign the work.