As was done in matgwinc some months ago

These changes improve the modeling of suspensions that have a
temperature gradient along the fibers.  Each stage is split into an
upper and a lower joint, which can have different material properties
assigned in the ifo structure.  The temperature of a joint is that of
the mass to which it is attached.  (However, temperature gradients are
not yet supported within the top or bottom stages.)  The average of
the upper and lower joint dissipation is used to compute the noise.

Removed normalization of sinc_sqr in dhdl because it is one at DC. Previously the normalization was set by the first frequency point, which is wrong if only high frequencies are considered

implement as __main__ for `gwinc.ifo`

also add to CI.  closes #48

This allows for specifying an order that makes sense without too much loss
of ease of expandability (just need to add an entry to a list).

This finally gets rid of the whole precompIFO function.  It does this
by breaking up precomp into various smaller functions, moving them
into the common IFO noise definition module, and using them
appropriately when needed.  This does result in some of the functions
being called multiple times in the full budget calculation for the
aLIGO-like budgets, but the cost should be minor given the convenience
of getting rid of the whole precomp thing, thereby opening up
everything to non-aLIGO-like configurations.

Even though some functions are calculated multiple times, we're still
orders of magnitude faster than matgwinc, so...  Should still revisit
down the line though.

closes #40

Define new Strain(nb.Calibration) and apply it to all canonical ifo budgets.

residual_gas_cavity() function updated to calculate residual gas noise of
a single evacuated cavity.

ifo ExcessGas Noise class modified to handle specific case
of Fabry-Perot Michelson.

Simplify newtonian noise calculations to return displacement noise for
simplified configuration: either single test mass or single arm.

new suspension_thermal() function for calculating thermal noise of a
single suspended optic.

ifo SuspensionThermal Noise class modified to handle specific case of
Fabry-Perot Michelson.

substrate_thermoelastic() function updated to calculate thermal noise
of a single suspended optic.

ifo SubstrateThermoElastic Noise class modified to handle specific
case of Fabry-Perot Michelson.

substrate_brownian() function updated to calculate thermal noise of a
single suspended optic.

ifo SubstrateBrownian Noise class modified to handle specific case of
Fabry-Perot Michelson.

substrate_thermorefractive() function updated to calculate thermal
noise of a single suspended optic.

ifo ITMThermoRefractive Noise class modified to handle specific case
of Fabry-Perot Michelson.

substrate_carrierdensity() function updated to calculate thermal noise
of a single suspended optic.

ifo ITMCarrierDensity Noise class modified to handle specific case of
Fabry-Perot Michelson.

new coating_brownian() and coating_thermooptic() functions for
calculating coating thermal noise spectrum for a single optic.

ifo CoatingBrownian and CoatingThermoOptic Noise classes modified to
handle specific case of Fabry-Perot Michelson.

This refactors the seismic noise calculation to be for a single suspension.
The turning of that into a seismic noise spectrum for a LIGO-like IFO is
moved to the Seismic(nb.Noise) class definition in the ifo module,
effectively separating fundamental seismic noise calculation from assumptions
of experimental configuration.

This replaces load_ifo(), and returns only the Budget class.

Any ifo.yaml found in a budget package is loaded as ifo attribute.  other
stuff can just be assumed to be provided as class attributes as well.

move load_struct logic into Struct.from_file() class method, and remove
load_struct function.

This allows for direct comparison between py and mat gwinc.

Fix suspension: silicon blade spring

2 um wavelength

Seismic and Newtonian parameters

CE1 config still needs some FC debugging