RBLQ: add pure forecasting case

[szokeb87]

Adds the following case:
If Q and B are zero matrices, RBLQ assumes that the user wants to solve a "pure forecasting problem" defined on p171 in Robustness (Hansen and Sargent, 2008).

It is assumed that the user sets up the problem in a way that "eliminates" F. This involves redefining the R and A matrices in equations (1) and (2) (of this lecture) as
R_adj = R + F'QF
A_adj = A - BF
Then, using R_adj, A_adj, C with Q=B=0 (zero matrices), RBLQ returns a K which is the worst-case with respect to the proposed policy F.

[coveralls]

Coverage decreased (-0.07%) to 94.767% when pulling 05cbb3c on szokeb87:master into 707f8aa on QuantEcon:master.

[coveralls]

Coverage decreased (-0.07%) to 94.767% when pulling 776a8bf on szokeb87:master into 707f8aa on QuantEcon:master.

[jstac]

@szokeb87 Many thanks.
@mmcky For background, @szokeb87 has added this functionality in response to

http://discourse.quantecon.org/t/robust-lqc-special-case/281

[mmcky]

Thanks @szokeb87. Can we please add this case to the test_robustlq suite?

[szokeb87]

@mmcky Done

[coveralls]

Coverage increased (+0.02%) to 94.855% when pulling 245670c on szokeb87:master into 707f8aa on QuantEcon:master.

[cc7768]

Seems like your name should be added here 😉

[szokeb87]

@cc7768 this was just the cherry on the cake :)

[coveralls]

Coverage increased (+0.02%) to 94.855% when pulling 2aa5e6c on szokeb87:master into 707f8aa on QuantEcon:master.

[mmcky]

Thanks @szokeb87 this is looking great.