add compat for statsmodels 0.7

[kleinschmidt]

We will tag 0.7 very soon (probably today ahahaha good one...) with changes to FunctionTerm that are breaking but do not directly affect anything that GLM itself uses. This PR should not be merged until 0.7 is tagged and CI can be run to ensure that nothing breaks in GLM, so I'll start this in draft mode and flip it out when it's time to check CI.

Update: 0.7 is now released. I've added bumped the patch version here and also fixed a bug where we were defining our own formula function instead of creating a method for StatsModels.formula.

Update to teh update: based on #500, I re-based this on a commit before #339 was merged, targeting a "backport" branch called v1.

[codecov-commenter]

Codecov Report

Patch coverage: 97.52% and project coverage change: +1.35 :tada:

Comparison is base (9dc4d6b) 87.39% compared to head (7f816d7) 88.75%.

:exclamation: Current head 7f816d7 differs from pull request most recent head ceae65f. Consider uploading reports for the commit ceae65f to get more accurate results

Additional details and impacted files

```diff @@ Coverage Diff @@ ## v1 #512 +/- ## ========================================== + Coverage 87.39% 88.75% +1.35% ========================================== Files 7 8 +1 Lines 952 1040 +88 ========================================== + Hits 832 923 +91 + Misses 120 117 -3 ``` | [Impacted Files](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats) | Coverage Δ | | |---|---|---| | [src/GLM.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL0dMTS5qbA==) | `50.00% <ø> (ø)` | | | [src/ftest.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL2Z0ZXN0Lmps) | `100.00% <ø> (ø)` | | | [src/glmfit.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL2dsbWZpdC5qbA==) | `81.79% <94.28%> (+0.77%)` | :arrow_up: | | [src/lm.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL2xtLmps) | `94.26% <97.43%> (+0.93%)` | :arrow_up: | | [src/deprecated.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL2RlcHJlY2F0ZWQuamw=) | `100.00% <100.00%> (ø)` | | | [src/linpred.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL2xpbnByZWQuamw=) | `91.17% <100.00%> (+5.56%)` | :arrow_up: | | [src/negbinfit.jl](https://codecov.io/gh/JuliaStats/GLM.jl/pull/512?src=pr&el=tree&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats#diff-c3JjL25lZ2JpbmZpdC5qbA==) | `82.66% <100.00%> (ø)` | | Help us with your feedback. Take ten seconds to tell us [how you rate us](https://about.codecov.io/nps?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats). Have a feature suggestion? [Share it here.](https://app.codecov.io/gh/feedback/?utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=JuliaStats)

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.

[kleinschmidt]

Okay, I've double checked at least the ubuntu runs to make sure that statsmodels 0.7 was used and all tests are passing so I think this is good to go!

[kleinschmidt]

nightly failures are due to the "evaluated to nonboolean" change for @test_broken