Update continuous benchmarking workflow

[pratikunterwegs]

This PR supersedes #201 and fixes #184. See lorenzwalthert/touchstone#125 for why this was necessary. The workflow uses a development version of {touchstone} that allows emojis to be shown in the benchmarking comment.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 2c00500 is merged into main:

•   :ballot_box_with_check:default: 40.5ms -> 41.6ms [-0.83%, +6.4%]
•   :ballot_box_with_check:default_ode: 10.9ms -> 11.1ms [-0.15%, +2.74%]
•   :ballot_box_with_check:default_ode_interventions: 72.7ms -> 72.5ms [-1.04%, +0.6%]
•   :ballot_box_with_check:default_ode_param_vec: 835ms -> 834ms [-0.73%, +0.45%]
•   :ballot_box_with_check:default_ode_paramvec_intervs: 6.28s -> 6.3s [-0.17%, +0.86%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[sbfnk]

If you want icons you need the dev version, see
https://github.com/epiforecasts/EpiNow2/blob/21d01669d9d71024279046afcb7be9aca4656dbf/.github/workflows/touchstone-receive.yaml#L51
and
https://github.com/epiforecasts/EpiNow2/blob/21d01669d9d71024279046afcb7be9aca4656dbf/.github/workflows/touchstone-receive.yaml#L55

[pratikunterwegs]

Thanks, will use!

[pratikunterwegs]

Hard resetting to eb439d8 as dev workflow fails when base branch has errors in touchstone script (i.e., outside the benchmarking) - these were fixed in this PR.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 4e2d96b is merged into main:

•   :ballot_box_with_check:default: 46.9ms -> 47ms [-2.79%, +3.5%]
•   :ballot_box_with_check:default_ode: 11ms -> 11.1ms [-0.92%, +1.72%]
• ❗🐌default_ode_interventions: 73.3ms -> 73.8ms [+0.27%, +1.33%]
•   :ballot_box_with_check:default_ode_param_vec: 846ms -> 847ms [-1.22%, +1.44%]
• ❗🐌default_ode_paramvec_intervs: 6.33s -> 6.39s [+0.3%, +1.66%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 932e67a is merged into main:

•   :ballot_box_with_check:default_ode: 11ms -> 10.9ms [-1.99%, +0.72%]
•   :ballot_box_with_check:default_ode_interventions: 73.3ms -> 73.2ms [-1.04%, +0.74%]
•   :ballot_box_with_check:default_ode_param_vec: 844ms -> 838ms [-1.48%, +0.1%]
•   :ballot_box_with_check:default_ode_paramvec_intervs: 6.33s -> 6.33s [-0.67%, +0.77%]
•   :ballot_box_with_check:ebola: 41.5ms -> 41.3ms [-3.58%, +2.33%]
  Further explanation regarding interpretation and methodology can be found in the documentation.